sepsis-3-get-data.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import libraries\n",
    "from __future__ import print_function\n",
    "\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import psycopg2\n",
    "import socket\n",
    "import sys\n",
    "import os\n",
    "import getpass\n",
    "\n",
    "from collections import OrderedDict\n",
    "\n",
    "from sepsis_utils import sepsis_utils as su\n",
    "from sepsis_utils import roc_utils as ru\n",
    "\n",
    "# path to output data files to - can be relative to the current folder\n",
    "data_path = 'data'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# colours for prettier plots\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import colorsys\n",
    "def gg_color_hue(n):\n",
    "    hues = np.linspace(15, 375, n)\n",
    "    hsv_tuples = [(x*1.0/360.0, 0.5, 0.8) for x in hues]\n",
    "    rgb_tuples = map(lambda x: colorsys.hsv_to_rgb(*x), hsv_tuples)\n",
    "    return rgb_tuples\n",
    "\n",
    "marker = ['v','o','d','^','s','o','+']\n",
    "ls = ['-','-','-','-','-','s','--','--']\n",
    "\n",
    "# plot settings\n",
    "%matplotlib inline\n",
    "plt.style.use('ggplot')\n",
    "font = {'size'   : 20}\n",
    "matplotlib.rc('font', **font)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# helper function for generating tables\n",
    "# this rolls back the cursor if it fails\n",
    "def execute_query_safely(sql, con):\n",
    "    cur = con.cursor()\n",
    "    \n",
    "    # try to execute the query\n",
    "    try:\n",
    "        cur.execute(sql)\n",
    "    except:\n",
    "        # if an exception, rollback, rethrow the exception - finally closes the connection\n",
    "        cur.execute('rollback;')\n",
    "        raise\n",
    "    finally:\n",
    "        cur.close()\n",
    "    \n",
    "    return"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using username alistairewj\n"
     ]
    }
   ],
   "source": [
    "sqluser = getpass.getuser()\n",
    "print('Using username {}'.format(sqluser))\n",
    "\n",
    "dbname = 'mimic'\n",
    "schema_name = 'mimiciii'\n",
    "query_schema = 'SET search_path to public,' + schema_name + ';'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Connected to postgres 10.0.4!\n"
     ]
    }
   ],
   "source": [
    "# Connect to local postgres version of mimic\n",
    "con = psycopg2.connect(dbname=dbname, user=sqluser)\n",
    "\n",
    "print('Connected to postgres {}.{}.{}!'.format(int(con.server_version/10000),\n",
    "                                              (con.server_version - int(con.server_version/10000)*10000)/100,\n",
    "                                              (con.server_version - int(con.server_version/100)*100)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running SQL code to generate tables. This may take some time.\n",
      "-- As the script is generating many tables, it may take some time.\n",
      "\n",
      "-- We assume the database and the search path are set correctly.\n",
      "-- You can set the search path as follows:\n",
      "-- SET SEARCH_PATH TO public,mimiciii;\n",
      "-- This will create tables on public and read tables from mimiciii\n",
      "\n",
      "BEGIN;\n",
      "-- ----------------------------- --\n",
      "-- ---------- STAGE 1 ---------- --\n",
      "-- ----------------------------- --\n",
      "\n",
      "Running query/tbls/abx-poe-list.sql ... done.\n",
      "Running query/tbls/abx-micro-prescription.sql ... done.\n",
      "Running query/tbls/suspicion-of-infection.sql ... done.\n",
      "\n",
      "-- blood cultures around ICU admission\n",
      "Running query/tbls/blood-culture-icu-admit.sql ... done.\n",
      "\n",
      "-- generate cohort\n",
      "Running query/tbls/cohort.sql ... done.\n",
      "\n",
      "-- generate sepsis definitions\n",
      "Running query/../mimic-code/concepts/sepsis/angus.sql ... done.\n",
      "Running query/../mimic-code/concepts/sepsis/martin.sql ... done.\n",
      "Running query/../mimic-code/concepts/sepsis/explicit.sql ... done.\n",
      "\n",
      "\n",
      "-- -- Generate the views which the severity scores are based on (at time of infection)\n",
      "-- \\i tbls/urine-output-infect-time.sql\n",
      "-- \\i tbls/vitals-infect-time.sql\n",
      "-- \\i tbls/gcs-infect-time.sql\n",
      "-- \\i tbls/labs-infect-time.sql\n",
      "-- \\i tbls/blood-gas-infect-time.sql\n",
      "-- \\i tbls/blood-gas-arterial-infect-time.sql\n",
      "-- \\i tbls/vaso-dur.sql\n",
      "\n",
      "-- Generate the views which the severity scores are based on (first 24 hours)\n",
      "Running query/../mimic-code/concepts/durations/vasopressor-durations.sql ... done.\n",
      "Running query/../mimic-code/concepts/durations/ventilation-durations.sql ... done.\n",
      "\n",
      "Running query/../mimic-code/concepts/firstday/urine-output-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/ventilation-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/vitals-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/gcs-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/labs-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/blood-gas-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/blood-gas-first-day-arterial.sql ... done.\n",
      "\n",
      "Running query/../mimic-code/concepts/echo-data.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/weight-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/firstday/height-first-day.sql ... done.\n",
      "Running query/../mimic-code/concepts/comorbidity/elixhauser-ahrq-v37-with-drg.sql ... done.\n",
      "\n",
      "-- ----------------------------- --\n",
      "-- ---------- STAGE 2 ---------- --\n",
      "-- ----------------------------- --\n",
      "\n",
      "-- Severity scores during the first 24 hours\n",
      "Running query/../mimic-code/concepts/severityscores/lods.sql ... done.\n",
      "Running query/../mimic-code/concepts/severityscores/mlods.sql ... done.\n",
      "Running query/../mimic-code/concepts/severityscores/sirs.sql ... done.\n",
      "Running query/../mimic-code/concepts/severityscores/qsofa.sql ... done.\n",
      "Running query/../mimic-code/concepts/severityscores/sofa.sql ... done.\n",
      "\n",
      "-- -- Severity scores at the time of suspected infection\n",
      "-- \\i tbls/sofa-si.sql\n",
      "-- \\i tbls/sirs-si.sql\n",
      "-- \\i tbls/lods-si.sql\n",
      "-- \\i tbls/qsofa-si.sql\n",
      "-- \\i tbls/mlods-si.sql\n",
      "\n",
      "-- -- Severity scores on admission\n",
      "-- \\i tbls/qsofa-admission.sql\n",
      "-- \\i tbls/blood-gas-admission.sql\n",
      "-- \\i tbls/sirs-admission.sql\n",
      "\n",
      "-- ----------------------------- --\n",
      "-- ---------- STAGE 3 ---------- --\n",
      "-- ----------------------------- --\n",
      "-- Some sepsis criteria require the severity scores (e.g. SIRS)\n",
      "Running query/tbls/sepsis_cdc_surveillance.sql ... done.\n",
      "Running query/tbls/sepsis_nqf_0500.sql ... done.\n",
      "\n",
      "-- Generate the final table\n",
      "Running query/tbls/sepsis3.sql ... done.\n",
      "\n",
      "COMMIT;\n"
     ]
    }
   ],
   "source": [
    "# check if the sepsis3_cohort table exists ... if not we must generate it\n",
    "query = \"\"\"\n",
    "SELECT EXISTS(SELECT 1 FROM information_schema.tables \n",
    "              WHERE table_catalog = '{}'\n",
    "              AND table_schema in ('public','{}')\n",
    "              AND table_name = 'sepsis3');\n",
    "\"\"\".format(dbname, schema_name)\n",
    "tbl_exists = pd.read_sql_query(query, con)\n",
    "tbl_exists = tbl_exists.loc[0,'exists']\n",
    "if tbl_exists:\n",
    "    print('Found the `sepsis3` table. Skipping generation of data in SQL.')\n",
    "else:\n",
    "    print('Running SQL code to generate tables. This may take some time.')\n",
    "    \n",
    "    # read through the \"make-tables.sql\" file in the sql subfolder\n",
    "    query_path = 'query'\n",
    "    \n",
    "    with open(os.path.join(query_path, 'make-tables.sql'), 'r') as fp:\n",
    "        for line in fp.readlines():\n",
    "            if len(line)<2:\n",
    "                print(line,end='')\n",
    "                continue\n",
    "            \n",
    "            if line[0:2] != '\\i':\n",
    "                print(line,end='')\n",
    "                continue\n",
    "                \n",
    "            # lines which begin with '\\i' call SQL files that generate tables\n",
    "            query_file = os.path.join(query_path, line[3:].replace('\\n',''))\n",
    "            print('Running {} ...'.format(query_file), end=' ')\n",
    "            with open(query_file, 'r') as fp_query:\n",
    "                query = ''.join(fp_query.readlines())\n",
    "            execute_query_safely(query_schema + query, con)\n",
    "            print('done.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# exclusion criteria:\n",
    "#   - less than 16 years old\n",
    "#   - never have any chartevents data (i.e. likely administrative error)\n",
    "#   - not cardiac surgery\n",
    "#   - suspected of infection\n",
    "#   - first ICU stay\n",
    "#   - not a CareVue patient (i.e. admitted 2008-2012)\n",
    "# these exclusion criteria are created in the sepsis3_cohort table\n",
    "query = query_schema + \"select * from sepsis3_cohort\"\n",
    "co = pd.read_sql_query(query,con)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exclusions - applied independently"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cohort - initial size: 61532 ICU stays\n",
      "  15056 (24.47%) - exclusion_secondarystay\n",
      "   8109 (13.18%) - exclusion_nonadult\n",
      "   7343 (11.93%) - exclusion_csurg\n",
      "  37912 (61.61%) - exclusion_carevue\n",
      "   7343 (11.93%) - exclusion_early_suspicion\n",
      "   3914 (6.36%) - exclusion_late_suspicion\n",
      "    491 (0.80%) - exclusion_bad_data\n",
      "Final cohort size: 11791.0 ICU stays (19.16%).\n"
     ]
    }
   ],
   "source": [
    "# print out the exclusions\n",
    "print('Cohort - initial size: {} ICU stays'.format(co.shape[0]))\n",
    "idxRem = np.zeros(co.shape[0])\n",
    "for c in co.columns:\n",
    "    if c.startswith('exclusion_'):\n",
    "        print('  {:5g} ({:2.2f}%) - {}'.format(np.sum(co[c]),np.mean(co[c])*100.0, c))\n",
    "        idxRem[co[c].values==1] = 1\n",
    "        \n",
    "print('Final cohort size: {} ICU stays ({:2.2f}%).'.format(co.shape[0] - np.sum(idxRem), (1-np.mean(idxRem))*100.0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exclusions - applied sequentially"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cohort - initial size: 61532 ICU stays\n",
      "   8109 (13.18%) - exclusion_nonadult\n",
      "  14826 (24.09%) - exclusion_secondarystay\n",
      "   5905 (9.60%) - exclusion_csurg\n",
      "  18909 (30.73%) - exclusion_carevue\n",
      "   1250 (2.03%) - exclusion_early_suspicion\n",
      "    724 (1.18%) - exclusion_late_suspicion\n",
      "     18 (0.03%) - exclusion_bad_data\n",
      "Final non-suspected cohort size: 11791 ICU stays (19.16%).\n",
      "\n",
      "=============================\n",
      "====== METAVISION ONLY ======\n",
      " Patients admitted 2008-2012 \n",
      "=============================\n",
      "\n",
      "Cohort - initial size: 23620 ICU stays\n",
      "      3 (0.01%) - exclusion_nonadult\n",
      "   7536 (31.91%) - exclusion_secondarystay\n",
      "   2298 (9.73%) - exclusion_csurg\n",
      "      0 (0.00%) - exclusion_carevue\n",
      "   1250 (5.29%) - exclusion_early_suspicion\n",
      "    724 (3.07%) - exclusion_late_suspicion\n",
      "     18 (0.08%) - exclusion_bad_data\n",
      "Final non-suspected cohort size: 11791/23620 ICU stays (49.92%).\n",
      "\n",
      "Of these patients...\n",
      "   7061 (59.88%) - suspected_of_infection_poe\n"
     ]
    }
   ],
   "source": [
    "# print out the exclusions *SEQUENTIALLY* - i.e. if already excluded, don't re-print\n",
    "print('Cohort - initial size: {} ICU stays'.format(co.shape[0]))\n",
    "\n",
    "COL_REM = ['exclusion_nonadult', 'exclusion_secondarystay',\n",
    "           'exclusion_csurg','exclusion_carevue',\n",
    "           'exclusion_early_suspicion', 'exclusion_late_suspicion', 'exclusion_bad_data']\n",
    "\n",
    "idxRem = np.zeros(co.shape[0])\n",
    "for c in COL_REM:\n",
    "    N_REM = np.sum( (idxRem == 0) & (co[c].values==1) )\n",
    "    print('  {:5g} ({:2.2f}%) - {}'.format(N_REM,N_REM*100.0/co.shape[0], c))\n",
    "    idxRem[co[c].values==1] = 1\n",
    "        \n",
    "print('Final non-suspected cohort size: {:g} ICU stays ({:2.2f}%).'.format(co.shape[0] - np.sum(idxRem), (1-np.mean(idxRem))*100.0))\n",
    "\n",
    "print()\n",
    "# define idxRem to exclude non-metavision admissions initially\n",
    "print('=============================')\n",
    "print('====== METAVISION ONLY ======')\n",
    "print(' Patients admitted 2008-2012 ')\n",
    "print('=============================')\n",
    "print()\n",
    "\n",
    "idxRem = co['dbsource']!='metavision'\n",
    "N = np.sum(~idxRem)\n",
    "N_IGNORE = np.sum(idxRem)\n",
    "\n",
    "# print out the exclusions *SEQUENTIALLY* - i.e. if already excluded, don't re-print\n",
    "print('Cohort - initial size: {} ICU stays'.format(N))\n",
    "\n",
    "for c in COL_REM:\n",
    "    N_REM = np.sum( (idxRem == 0) & (co[c].values==1) )\n",
    "    print('  {:5g} ({:2.2f}%) - {}'.format(N_REM,N_REM*100.0/N, c))\n",
    "    idxRem[co[c].values==1] = True\n",
    "        \n",
    "print('Final non-suspected cohort size: {}/{} ICU stays ({:2.2f}%).'.format(\n",
    "        np.sum(~idxRem), N, np.sum(~idxRem)*100.0/N))\n",
    "print('')\n",
    "\n",
    "print('Of these patients...')\n",
    "c='suspected_of_infection_poe'\n",
    "N_REM = np.sum( (~idxRem) & (co[c].values==1) )\n",
    "print('  {:5g} ({:2.2f}%) - {}'.format(N_REM, N_REM*100.0/ np.sum(~idxRem), c))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Histogram of time of suspected infection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1IAAAGKCAYAAAAG1HBsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3XdUVMfbwPEvVRDsIuhaUNEk9tgwxg6KsSIqRqOxxN6D\nsaSJsSeRoib5aewxBtFoLFhQBFEBS6Jii8aGihVEUIogsO8fnr0vK7uwKEWT53MOJ3jvzNy5l4Xs\nszPzjJFarVYjhBBCCCGEEMJgxkXdASGEEEIIIYR400ggJYQQQgghhBB5JIGUEEIIIYQQQuSRBFJC\nCCGEEEIIkUcSSAkhhBBCCCFEHkkgJYQQQgghhBB5JIGUEELko0OHDmFkZIS1tTUxMTFF3R2RT7p3\n746RkRGjR4/Ol/aioqIwMjLCyMiIWbNm5Uubr8Le3l7pT9av+Pj4ou6aQP/PJyoqqqi7JsR/mgRS\nQvxHDRkyROf/mHP6sre3z7XdTZs20a1bN1QqFcWKFUOlUtGtWzc2bdqU5z7+888/fPnllzRp0gQb\nGxulPUdHR6ZMmUJoaKjeunFxcezbt48FCxbQu3dvqlWrlqf7eBkZGRmMGzcOgIkTJ2JjY5OtzMGD\nBw1+3qVLl87xeo8fPyY0NBQvLy/69+9P7dq1MTY2Vurnp1u3brF69WrGjRvH+++/T40aNShVqhTm\n5ubY2NjQunVrPD09uXnzZq5t5Ve/MzMz+fHHH2nUqBGWlpaULl2ajh07EhQUlGtdtVpNmzZtMDIy\nwtfXN9fy33zzDQArVqzg5MmTBvfxv+7hw4f8+uuvDB8+nMaNG1OmTBnMzMwoW7YszZs3Z/r06Vy9\nevWVrtGrVy+t35uDBw++cr/Pnj3LokWL6NmzJzVr1qR48eJYWFigUqno2rUrK1euJCUl5aXbv3Xr\nFiVLlizwv0lCiAKmFkL8Jw0ePFgN5Omrc+fOettLTExUf/DBBznW79KlizopKSnXvmVkZKi//PJL\ntZmZWY7ttW3bVm8bpUuX1luvWrVqL/HEcrd69Wo1oC5ZsqT64cOHOsuEhIQY/LxLlSqV4/UaNWqU\nY/38NGfOHIP6bGlpqV66dGmh9HvgwIE66xsZGalXrFiRY90VK1aoAXXjxo3V6enpBl2vV69eakDd\nqVMng/uoz/Xr15X+enp6vnJ7r6patWpqQN20aVP12bNnla+MjIyXbnPFihVqU1PTXF8zZmZm6gUL\nFrzUNfz8/LK1FxIS8tJ9VqvV6rZt2xr0Wq9Ro4b6+PHjL3WNzp075+lv0qVLl5SfyZgxY5R6169f\nf6nrCyHyh2lOQZYQ4t9r3rx5fPbZZ7mWGzduHIcOHQJg6NChesv179+fPXv2ANCiRQs+/fRTatSo\nwbVr1/Dx8eHo0aPs3r2b/v37s337dr3tZGZmMmTIENavXw9A48aN+fjjj2nUqBGlSpXi0aNHnDt3\njp07d+bYb7VarXxfvnx5mjRpQlhYGImJibne88vIyMhgzpw5AIwYMYKyZcvmWmfu3Ln07NlT73kT\nE5Mc62e9x1KlSvHuu+9y8eJF7t27Z2CvDWdkZESdOnVo3bo1DRs2pFKlStjZ2ZGens6dO3fYvn07\nmzdvJiUlhQkTJmBtbc2QIUMKrN+bN2/m119/BWDQoEEMGzaMJ0+eMGvWLE6ePMn48ePp3LkzlStX\nzlY3JiaG6dOnY2xszPLly3N9zhrTpk3jjz/+YN++fURERPDee+8Z3N8X2dvbaz2H14WVlRX16tXL\nl7bu3btHeno6JiYmODs706lTJxo1akSZMmWIiYlh165dLFu2jLS0ND7//HMAZsyYYXD7MTExTJgw\nAQBbW1vu37+fL/2Ojo4GoFy5cri5udG2bVuqV69OsWLFuHTpEsuWLePw4cNcu3aNjh07cvToUd5+\n+22D21+7di179+7NU59r166tfF+hQoW83ZAQouAUbRwnhHidxcTEqM3NzdWAumzZsuqnT5/qLLd5\n82blE1JnZ2d1Wlqa1vnU1FS1k5OTUmbLli16r+nl5aWUmzt3rjozM1Nv2dTUVL3n5s6dq96yZYs6\nKipKOab51L0gRqT8/f2VfkdGRuotl3VEas2aNa90zcWLF6t/++039T///KM8p6yfpuenZ8+e5Vrm\nyJEjygiEnZ2d3tGM/Oh3p06d1IDayclJ6/ijR4+U0cj58+frrPvRRx+pAfWECRMMulZWtWrVUgPq\nPn365Lnu60zzu5HTKG9eeXt7q6dMmaK+ffu23jKHDx9WW1hYqAG1ubm5+ubNmwa3369fP2WkPOsI\n+6uOSHXp0kW9fv16vX9fMjMz1RMmTDBopP5Fd+7cUZcpU0YNqDdu3PhSo+Senp4yIiXEa0ICKSGE\nXr6+vsr/sCdOnKi3XJMmTdSA2tjYWH316lWdZa5evao2NjZWpg/pcvv2bbWlpaUaUH/88cf5cg9Z\nFWQgpQkUGzZsmGO5/AykdCmoQMpQmgAHUJ87d87genntd7ly5dSAevXq1dnODRo0SG+ws3//fjWg\nrlSpkjohIcHg/mnMnj1bDahNTU3VDx48yHP911VBBFKG+vTTT5Wf/eLFiw2qs23bNjWgtrKyUkdF\nReVrIGWI1NRUtZ2dnRpQm5iY6J3K+yJXV1c1oO7WrZtarVZLICXEG06STQgh9FqzZo3y/bBhw3SW\niYqK4q+//gKgQ4cO1KhRQ2e5GjVq0K5dOwD+/PNPndmmfv75Z1JSUl6bTGaGunPnDiEhIQC4u7sX\ncW+KVsmSJZXvnz59WmDX0WSTq1SpUrZzmmMvZpx7+vQpY8aMAWDJkiVafTVUv379AEhPT8ff3z/P\n9TVyy9qn6/zp06cZPHgw1apVo1ixYlSoUIEePXrkmHTlTdChQwfl+3/++SfX8vHx8crPce7cuVSr\nVq3A+qaPubk577//PvB8Wu+1a9dyrePv78+2bduwtrbmp59+KuguCiEKgQRSQgidTp48SWRkJPB8\nnVLDhg11ltOsnwKUQEmf9u3b66yn4efnB0DDhg2pXr26cjw2NpbLly8TFxdncP8L0+7du8nMzASg\nZcuWeaqbnJzM1atXuXHjBsnJyQXRvUJz//59Dhw4AICxsbHWuo78VqpUKQCd66ru3r0LkC3r4bx5\n87hy5Qpdu3ald+/eL3Xd2rVrU65cOQACAgJeqo2X8fPPP9O8eXN++eUXbt68SVpaGjExMezcuZP2\n7duzbNmyQutLfktLS1O+N2S92qeffsrdu3dp1qwZEydOLMiu5Sgv/Y6NjVXWc82fP58qVaoUaN+E\nEIVDAikhhE6rV69Wvtc3GgVw/vx55fs6derk2GbW8xcuXNA6FxcXp3wa3ahRI9LT0/n++++pUaMG\nNjY2yhvYmjVr4unpyePHj/N0PwVJMyJgbGxM06ZNDa43Y8YMSpYsiYODA/b29pQoUYJGjRrx7bff\nvlb3l5Pk5GSuXLnCDz/8QIsWLXj06BEAI0eOpESJEgV23caNGwOwceNGreNPnjxRAhxNGYC///6b\n7777juLFi/PDDz+80rUdHR0BOHz4MBkZGa/UliH27dvHmDFjqF27NitWrODYsWOEhYXx1VdfYW5u\njlqtZtKkSQaN5ryONKO5kPvfkMDAQNauXYupqSkrV67E2Lho3sakpaURHh4OgJmZGQ4ODjmWnzBh\nAjExMbRo0ULZIkEI8eaTQEoIkU1qaiq//fYbABYWFgwYMEBv2az7BuX2KWvW8y/uN3Tu3Dnle2tr\na5ydnZk2bRrXr1/XKnft2jVmz55NkyZNuHLlSu43Uwg0Uxtr166NtbW1wfXu37+v9UY8MzOTyMhI\nZsyYQf369Tlx4kS+9zU/zJo1S5l2ZmVlRa1atZgwYYIyXbNHjx589913BdoHTQbJvXv3MmzYMEJD\nQwkICMDZ2Zm4uDiKFSvGwIEDgedZAkePHk1aWhqenp6vvGdPs2bNgOdB5IsfCBSEiIgIOnbsyF9/\n/cXw4cNp3rw5LVu2ZM6cOaxYsQJ4/sb+TRyVio6OZu3atQBYWlri6uqqt+yTJ08YOXIkAFOnTqVB\ngwaF0UWdfvrpJx4+fAg836w5pw8NduzYwcaNGzEzM2PFihVFFvwJIfKf/DYLIbLZtm2bMrLg6upK\nmTJl9JZ98uSJ8n1uQUTW81nrAVrT9lauXEloaCjvvPMO27dvJyEhgeTkZA4dOkSbNm0AuHLlCj16\n9HilTTHzg1qtVjYUtbW1zbW8kZER77//PosXL+bUqVM8fvyYtLQ0oqOj2bBhA40aNQKeB5ouLi5v\n1CiDvb09O3fuZNu2bQU6GgUwYMAA+vbtCzxfy9euXTu6d+/O8ePHMTIyYvHixUrgvnr1ag4dOkT9\n+vXx8PB45Wtn/TkXRjBvYWHBunXrKFasWLZzAwcOxM7ODuCNWyuVkZHB0KFDlS0JPvvssxx/h6ZP\nn87NmzepVasWM2fOLKxuZnPhwgW+/vprAIoVK8bcuXP1ls26nmvGjBn5llpeCPF6kEBKCJGNIUkm\nNLIGMubm5jmWzfpG8MUAKOv+Tk+fPqV69eqEhYXRo0cPSpYsiaWlJa1btyYoKEhZa/X333+zfPny\n3G+oAMXFxSlrJQzZO6pNmzYcOXKEiRMn0qhRI0qUKIGZmRkqlYoBAwZw/PhxPvzwQwAePXpUpGtA\n9Bk7dixnz57l7NmzHD9+nI0bN+Lu7s6NGzcYNWoU//vf/wqlH35+fvj6+tKgQQOKFStGiRIl6NCh\nA3v27GHUqFHA872Gpk2bhpGREcuXL8fU9Pn2iX///Tf9+vWjQoUKFCtWjNq1a+Pp6WlQYJ7151wQ\ne3a9yNnZWW+AYWxsTJMmTQAMSnjwOvHw8CAoKAh4Pl1SE5zoEhoaqoy4/fzzz1hYWBRKH18UGxtL\n9+7dlb9X33//Pe+8847e8h4eHty5c4e33nqLL7/8srC6KYQoJLIhrxBCS3R0NPv37wegatWqODk5\n5Vg+6xuarIuvdUlNTVW+t7S01NsOwJw5c3SOhJmZmeHt7c27774LwIYNG5g8eXKO1y1IWRNE5DRy\np2FkZJTjeTMzM1atWkVoaCh3794lMDCQqKioV56Olp8qVKigtSlos2bN6NevH4MGDaJ3796MGzeO\nixcvsmTJkgLth4mJCZMmTWLSpEl6y0yZMoW4uDhGjRqlbKB77NgxnJ2dSUxMpFSpUtSqVYuLFy8y\ne/ZsDhw4QHBwcI4fCmQNpJKSkvLvhvTIbbNXTX8KY13d9evX9d6zubm5wQlGFi5cqLw+7O3t2bJl\nC2ZmZjrLpqSkMHz4cNRqNZ988kmuSW0KypMnT+jSpYsSsI4YMUJJIKHLvn37WLNmDUZGRqxYsULn\niKIQ4s0mI1JCCC3r1q1TMtANGTIk1/n8WadwZR1V0iXr+Renfr347y5duuhtp1GjRkqK61OnTvHs\n2bMcr1uQNCMckH/pvosXL66k2QbdGQ5fR926dVPeWC5dulQriUBRCA4OZv369dja2rJw4ULg+VTM\nwYMHk5iYiJubG/fv3+fcuXNERkZia2tLWFgY3t7eObabddRK35v//GRlZZXjec3vqOb3tiANHTqU\n+vXr6/zq1KmTQW0sWbKEzz//HIDKlSsTHByMSqXSW/6rr77iypUr2NnZsWjRony5j7xKSkqia9eu\nyrrFgQMH5rgmLTExUVnPNXLkSFq3bl0o/RRCFC4ZkRJCaNEs/DYyMlIW9OekatWqyve3bt3KMWvd\nrVu3dNYDtPaCKVGiRK6jO1WrVuXOnTtkZGQQFxdn0PqkgpA1xbZm8Xl+yPrJfmFMH8svbm5ueHl5\nAc8z6mVNeV+YUlNTGT16NAA+Pj7KzyksLIxLly5hZmbGjz/+qIwS1K1bl6lTp/LZZ5+xevVqZsyY\nobftrD/nF1Osi5wtW7ZMGUGsWLEiwcHBWlsdvCgxMRFfX18AOnbsyN69e3WWy5qUJjg4WPmd6dy5\n8yv/jFJSUujevTuHDx8Gnu8ltnbt2hw/ZFq/fj03btzAxMSEd999N1t2yRclJSUpZcqUKYOLi8sr\n9VkIUTgkkBJCKA4dOqQsnm/fvr1B08nq1q2rfH/hwgV69eqlt2zWDGcvpjmuVasWxYoVIzU11aCU\n0lnLZB0VKmyWlpbY2NgQExOTr/tcvS73l1fly5dXvte16XJhmTdvHpcvX6Zjx470799fOX7q1CkA\natasqSRp0NBssHr58mUSExP1Jk/RJGIBimQz2KJ08ODBl667cuVKxo4dCzyfHhocHEytWrVyrJOe\nnq6MtK1fv57169fnep05c+Yo3586dUpJ4PIynj59Ss+ePZXRVTc3N3799ddc943STGPOyMhQAvqc\nxMbGKq/Thg0bSiAlxBtCpvYJIRSG7h2VlSaLHuT+JivrVK8Xp7qYmJgox5KTk7lz547edrJmyite\nvLhBa5MKkiYT1+XLl1Gr1fnSpmYzZCDHaU+vm9u3byvfF3TmPn0uXbrEt99+i4WFBT/99JPWuYSE\nBOD/N/TNKuvIhaacLhcvXlS+lyxshlm7di0jR45ErVZjY2NDcHBwrmu/ilpqaiq9evVS1oz27NmT\njRs3vlEfbAghCpYEUkII4PkUmt9//x14/ibTzc3NoHrVq1dXNj4NDg7Wmzns2rVrSqDVpEkTndN5\n3N3dle83bdqk95oHDhxQRn9at25d5PuyaDZojY+P59KlS6/cXkxMjHL/xsbGRTY97mX4+fkp3xfV\nPj+jRo0iLS2NL7/8MttGqZoAKus0U42se5vpCrQ0jh07BoCDg4PWCJzQ7ddff+WTTz5BrVZTvnx5\nDhw4oDWSnZPSpUujVqtz/Ro8eLBSJyQkRDn+sqNRaWlp9OnTR5lK2K1bNzZt2mTwmrjJkycb1G+N\natWqKcdOnz79Un0WQhQ+CaSEEAD4+/sr2bj69++fLateTjTrSTIzMxk9enS25A/Pnj1j1KhRyhQd\nfetPBg0apEyVmj17NpcvX85WJi4uTisl+Pjx4w3uZ0HJOg3n6NGjess9evRI+XRbn4cPH9KrVy8l\nA9uAAQO0MuQVhKwb7M6aNSvb+djYWDZu3JjraNu6detYtWoV8DyD20cffVQQ3c3RmjVrlD3Ipk2b\nlu28Jui/c+cOwcHBWuc008Zq1aqld1pfYmIi58+fB5DpVwbw9/dnyJAhZGZmUq5cOYKCgqhfv36R\n9sne3l55veuafpqenk6/fv0ICAgAnie+2bJlS67bO7wuDh48qNxfUWU4FOK/QsanhRBA3vaOelHf\nvn3p0qULu3fvZv/+/bRp0wYPDw+qV6/O9evX8fb2VgKMrl270qdPH53tWFhYsGzZMrp3786jR49w\ndHTks88+o127dpiZmfHnn3/y3XffKW9++vfvT7du3XS2dfr06Wyf7GqyBiYmJipJNTQcHBxo1apV\nnu5bo1WrVlSoUIEHDx6wf/9+hgwZorNcQkICnTp14u2338bV1ZWmTZuiUqmwsLDgwYMHHDp0iJ9/\n/pmYmBjg+ToeTeIGXa5cucKRI0e0jmVNTPHiPdrZ2dG5c+c8319iYiL9+/dn+vTp9O7dm+bNm1O1\nalWsrKx4/Pgx58+fZ9OmTVpTNxcsWEDNmjULtd+xsbFMnToVeJ7UQNcb35YtW/LWW29x6dIlPvzw\nQ77//nscHBzYsmULv/76KwCffPKJ3msEBwcr69d69+6da5/+y3bs2MHAgQPJyMjAzMwMHx8fTExM\nOHfunN46ZcqUKfKprAMGDGDbtm3A8zWgs2fPznVj7OrVq+eaXVEI8S+kFkL85126dEkNqAF1vXr1\nXqqNJ0+eqDt16qS0o+vLxcVFnZiYmGtbGzduVJcoUSLHtgYNGqR++vSp3jY8PT1zrP/i1+DBg1/q\nvjU8PDzUgLp48eLqJ0+e6Cxz/fp1g/vTrl079a1bt3K85po1a/J0j23bts31WXl6er5Sv0uXLq1e\nuXJlofT7RR9//LEaUA8dOjTHchEREWpra2ud12rTpo06LS1Nb90+ffqoAXW1atXUmZmZBvVLl6zP\nNLdnrut8VoMHD1bKvqxq1arl6VkbImu/Cvr3MOu1QkJCcixboUIFNaA2NzdXP3z4MNv5vPbZkGvq\no6lfrVo1g+tk/X29fv16tvMhISF5/t0RQrwcmdonhHil0SgNa2trAgMD8fPz44MPPsDOzg4zMzPs\n7Oz44IMP2LhxI3v37jXoU9t+/fpx/vx5ZsyYQf369SlVqhQWFhZUq1aNAQMGEBISwi+//PJabXA5\nfvx4TExMSE5OZsuWLTrLVKpUiS1btjBt2jTat2+Pg4MDZcqUwdTUlNKlS1OvXj2GDx9OcHAwISEh\nVK5cuZDvQreqVasSFhbGN998Q9euXalbty42NjaYmppibW1NjRo1cHV15aeffuLatWs5jugUlIMH\nD/LLL79Qrlw5vv/++xzLtmjRgqNHj9K3b1/KlSuHmZkZDg4OeHp6snfvXr3rYOLj49m5cycA48aN\ny3VzZfH6uXLlCg8ePACejzxm3VxZCCHyykitzqcUU0II8R/30Ucf8dtvv9GyZUvCwsKKujsiny1Z\nsoRJkyZRpkwZrl+/nmNCijeNvb09N27coG3btq+U4vx1t2LFCkaOHEmxYsW4evVqkU8jfBmzZs3i\nm2++AZ7vn2XINhVCiIIhI1JCCJFPvvnmG8zMzAgPD2ffvn1F3R2Rj54+fcrChQsBmD59+r8qiMoq\nKSmJc+fOKV+aBDH/Fpp1fKNHj36jgqh//vlH+ZloRtSEEEVPRqSEECIfTZ06lUWLFtGiRQsiIiKK\nujsin/j4+ODh4UGNGjU4f/48FhYWRd2lfKUZkXrRo0ePtPbXetNVqlSJhIQErl27hq2tbVF3x2D6\nfj4yIiVE0ZKsfUIIkY88PT2xtrZGrVYTExODjY1NUXdJ5AMLCws8PT1xcXH51wVR/yU5bfQthBB5\nJSNSQgghhBBCCJFHskZKCCGEEEIIIfJIAikhhBBCCCGEyKP/3Bqp/J4fXb58eWJjY/O1TaGfPO/C\nJ8+8cMnzLnzyzAuXPO/CJ8+8cMnzLnz5/cwrVapkUDkZkRJCCCGEEEKIPJJASgghhBBCCCHySAIp\nIYQQQgghhMgjCaSEEEIIIYQQIo8kkBJCCCGEEEKIPJJASgghhBBCCCHySAIpIYQQQgghhMgjCaSE\nEEIIIYQQIo/+cxvyGkqtVpOamkpGRgYARkZGOsvdv3+f1NTUwuzaf5o878Inz7xw5fV5q9VqAExM\nTChWrJjev1VCCCGEyF8SSOmgVqtJSkrCwsIi1zcmpqammJiYFGLv/tvkeRc+eeaF62Wet1qtJiMj\ng6SkJKysrCSYEkIIIQqBTO3TITU1FQsLC0xNTeUNiRDitWdkZISpqSkWFhakpaUVdXeEEEKI/wQJ\npHTIyMiQT+CFEG8cExMT0tPTi7obQgghxH+CBFJ6yEiUEOJNI3+3hBBCiMIja6R0kDcjQog3lfz9\nyl3GiB7aB/4IL5qOCCGEeKPJiJQQQgghhBBC5JEEUkIIIYQQQgiRRxJIiXzl7++PSqXC39+/qLui\nuHXrFiqVismTJxd1VwDw8vJCpVIRHi7TiYQQQggh3lQSSAm9VCpVtq/q1avj6OjIpEmTuHz5clF3\n8V9FpVLRp0+fou7Ga2/x4sXK6/HKlSt6y2mC+he/ateujYuLC0uWLCElJUWrzoMHD6hfvz61atXi\n+vXrOtvduXMnKpWK7t27Kxt2CyGEEOK/R5JNiFx5eHgo3z9+/JjTp0/z+++/s3v3bv744w/q1atX\nhL0T/yVqtRo/Pz+MjIxQq9X89ttvzJw5M8c6derUoXPnzgBkZmYSExPD/v37+fbbbzl48CCbN29W\ntjuoUKEC3377LSNGjGDixIls27ZNayuEu3fvMmPGDKysrFi6dKlskyCEEEL8h+UpkHr48CH+/v5E\nRkby5MkTypQpQ7NmzejTpw/W1tYGtXHmzBlOnz5NVFQUUVFRJCYm8tZbbzFnzpwc60VHR7Np0yYu\nXLhASkoK5cuX5/3338fV1RVzc/O83IbIoylTpmQ79tVXX7FmzRpWrlyJr69vEfRKvAm8vLzw9vbm\n6NGjVKlS5ZXbCw0N5datW7i7uytB0IwZM3L8G1C3bt1sr+GEhAScnZ05duwYx44do2XLlsq5Ll26\n0LdvXzZv3sySJUv49NNPgedB3Keffkp8fDzfffcd9vb2r3w/QgghhHhzGTy17969e8yYMYODBw9S\ns2ZNunbtSoUKFdi9ezdfffUVT548MaidvXv3EhAQwKVLlyhTpoxBdS5fvsznn3/OiRMnqF+/Ph98\n8AHFixfn999/Z86cOTx79szQ2xD5pE2bNsDz4NpQZ86cYcSIETRo0IDq1avTvHlzPv/8c+7fv6+z\n/P379/niiy9wdHTE3t6e+vXrM3ToUM6cOaOzfGJiIrNmzaJJkybUqFGDNm3asHz5cjIzM/N+gzx/\n0z548GClv02bNmXo0KEcOnRIKZPbmjBDputp2gCIiIjQmobm5eUFQHh4uNa/X+To6Iijo6POdv39\n/QkJCaFPnz68/fbbyrU0rly5wuTJk2natCn29vY0bNiQcePG5Thtrqhs2LABgAEDBtCrVy/i4uLY\ns2dPntspVaoUjRo1AnS/hufMmUPlypXx9fUlMjISgFWrVnH48GE6duzIRx999Ap3IYQQQoh/A4NH\npFatWkVCQgJDhw7lgw8+UI6vW7eOXbt24efnx8iRI3Ntx9XVlf79+6NSqYiNjWX8+PE5ls/MzOSn\nn34iNTWVadOm0bRpU+W4j48Px44dY9euXbi6uhp6KyIfHDlyBICGDRsaVH7//v2MHDkStVpN165d\nqVy5MmfOnOGXX35h3759/PHHH1StWlUpf/PmTXr16sW9e/eUkcc7d+4QEBBAUFAQP//8Mx07dlTK\np6am0q9fP06fPk2dOnVwc3MjISGBxYsXc/To0Tzf36JFi/Dx8cHKygoXFxcqVarE/fv3+fPPP9m6\ndasSSOaHunXr4uHhgbe3N5UrV8bd3V059957771y+7t27eLgwYO0b9+eQYMGER0drZwLCQlh+PDh\npKen07FzW4mlAAAgAElEQVRjR+zt7bl79y579uzhwIEDbN68mXffffeV+5AfNFPyatSoQbNmzShR\nogTLly9nw4YN9OzZM09taaaoGhsb65yaWqJECXx9fXF3d2fChAksXbqUBQsWUL58eRYtWpRftySE\nEEKIN5hBgdS9e/eIjIzExsYGFxcXrXPu7u4EBQVx+PBhPv74YywsLHJsq3bt2nnq4IULF7h9+zbv\nvPOOEkQBGBsbM3DgQI4dO8b+/fvp2bOnbERZQLKOgjx58oTIyEhOnDiBs7Mzo0ePzrV+UlISkydP\nJj09nd9//11r5OTHH39k/vz5TJ8+HT8/P+X4jBkzuHfvHtOmTWPSpEnK8WHDhtGzZ08mT57M8ePH\nsbKyAmD58uWcPn2aLl26sHz5coyNnw+2jh8/XivwN0RoaCg+Pj5UrVqVrVu3UrFiRa3zd+7cyVN7\nualXrx716tXD29ubKlWq6JxK+SqCg4NZv3497du31zoeHx/PuHHjsLS0ZOvWrVq/mxcvXqR79+58\n9tlnHDhwIF/787L8/f159uyZEmi+/fbbNGjQgPDwcK5fv0716tV11jt//rzyGtaskTpw4ACPHz9m\nzpw5euu99957jBw5kmXLltGrVy9SU1P53//+R/ny5QvmBoUQQgjxRjFoat/58+eB56MPmjeoGpaW\nlrz99tukpqYWSBa3c+fOASjTcLKytbWlYsWKxMTE6J0eJl6dt7e38rVixQqOHz9OrVq1cHV1NWht\nXGBgIPHx8fTo0SPb9LNRo0ZRpUoVDh06xO3bt4HngUpoaCgqlYqxY8dqlW/WrBk9e/YkPj6e3bt3\nK8f9/f0xNjbmyy+/1HqNVq1alWHDhuXpflevXg3AzJkzswVRAJUqVcpTe0XNxcUlWxAF8Pvvv5OQ\nkMBnn32W7QOOt99+mwEDBnDu3DkuXbpUWF3VS5NkwtjYWGuqpLu7u5J0Qp8LFy4or19fX182bNjA\nvXv36NSpE++//36O150yZQqWlpakpqbSo0cPOnXqlG/3JIQQQog3m0GBlOYTeF1vKgHs7OyA5xmt\n8ltu19YcL4hri+du376tfF2+fJmAgADKly/P+PHjWbhwYa71z549C6DzTaupqakSXGmCZs1/HR0d\nMTMzy1anVatWWuUSExOJiorCzs5OZwKAvE6PO3nyJEZGRrRr1y5P9V5Xuj6EAPjrr7+A54GGl5dX\ntq9r164BGPwBia5U497e3gC0aNEi2zl96710OXLkCFFRUbRp00brb4Em2cymTZv0rpXs27ev1mv4\n9OnTLF26lNDQULp166a8PnX56aeflBTpERERxMXFGdxnIYQQQvy7GTS1Lzk5GYDixYvrPK85npSU\nlE/dyr9rBwUFERQUBMDChQsNmpZz//59TE0NT2iYl7Jvoqz3V7JkSZo1a8batWtp1KgR//vf/xg6\ndKiSwECTDtrExESpl5iYCDwPenU9K00gnpiYiKmpqfIzt7Oz01le80b6xfIVKlTIsbyxsbFBP6vH\njx9TunRpSpQokWtZXff7IiMjI61zmhEzXXVeLPvidfTdg2Zaa9Zzmjr6nmN8fDzw/wkc9ElKSjLo\nuX322WfZjoWHhxMeHs6IESMoVaqU1rmWLVsa/LujGXH68MMPterY2NjQqVMnZe1c9+7dlXP6nlnF\nihVxd3cnLS2NKVOmsHDhQp3JQv766y+WLl1K1apV6du3L15eXnz++eesWrXKoD6/ilf5m1KsWDGZ\nfpiLF+cvmJqayjMrRPK8C58888Ilz7vwFdUz/3dHAICzszPOzs7Kv2NjY3Otk5qaavD+MKampqSn\np790/94Euu7PysqKmjVrcvbsWU6fPo2trS2AskFpRkaGUk8z/e/evXs627p3757SZnp6uhIc6ypv\namqqjD6+WP7Bgwc629eUz8zMNOhnVbJkSR49esSTJ0+wtLTMsaxarQYgLS0tW9sJCQlKmaznNFkE\nsz6jrO3p6mNO19Fcq2TJklrnND8Lffet+bns37+fOnXq6L1HQ1/jmjThWWVmZhIeHs4nn3yiM/25\nIe0+fPiQvXv3AjB69Gi96/J++eUXrfVwud2/JlHKqVOnsp1PTk5m3LhxZGZmsmTJEpo0acKRI0cI\nCAjA39+f3r1759rvl/Wqf1NSU1MN+jsn/l96ero8s0JUvnx5ed6FTJ554ZLnXfjy+5kbuozDoKl9\nmjeqmk/+X6Q5rln4n5+K8toiZ5pAIbf04pqsaOHh4dnOpaenc/z4cQDq16+vVf7EiRM631CGhYVp\nlbe2tsbe3p579+4RFRWVrXxERIQht6No3LgxarWagwcP5lpWM8qiKwGFJm22oYyNjZU3/3m5zvXr\n13n8+HGergXQpEkTAOX5v642b95MWloaDRo0oH///jq/ypUrx+HDh7l586bB7eb0+v3mm2+4fv06\nY8eOpVmzZhgbG+Pr64uVlRVff/11viccEUIIIcSbx6BAShOV6VuHpBlR0LeO6VXkdm3N8YK4ttBv\n79693Lx5EzMzM61sirp07tyZ0qVLs337dmVdjsbKlSu5efMmrVu3VqYHVqpUiTZt2nDr1i1Wrlyp\nVf6vv/5i27ZtlC5dWmv0oV+/fmRmZjJv3jytN8Y3b95UkkcYSpOcYvbs2Tpfd1mPaRKw/PHHH8pa\nGoBHjx4xb968PF23TJkyet+gOzg4UKJECfbt26f1iUtKSgozZ87M03U03N3dKVWqFN7e3pw6dSrb\nec1oUlHTTD2cP38+ixYt0vk1cODAXJNOZJWRkaFM0XtxDd2BAwf49ddfs23kW7VqVTw9PUlISGDK\nlCnKKKEQQggh/psMmtpXt25d4Pkn7JmZmVpZ0VJSUrh48SLFihWjVq1a+d7BevXqsXXrVk6fPk2v\nXr20zt2/f5+7d+9iY2OjTC0T+S9rUoDk5GT++ecfQkJCgOdpym1sbHKsb2Vlhbe3N6NGjaJPnz50\n69aNSpUqcfbsWUJDQ6lQoQLffvutVp2FCxfi6urKnDlzCA0NpUGDBty5c4ddu3ZhbGyMt7e3VsbA\nUaNGERgYyO7du3FxcaFdu3YkJCQQEBCAo6Mj+/btM/h+27Zty6RJk1i8eDHt2rVT9pGKjY3l+PHj\nNG7cGF9fX+B55shevXqxZcsWOnbsiJOTE4mJiQQHB+Po6KgkxDBEq1at2L59O4MHD6Z+/fqYmprS\nokULWrRogZmZGZ988gm+vr64uLjQuXNn0tPTOXz4MLa2tso6s7woW7Ysy5cvZ/jw4XTv3p1WrVpR\nu3ZtjIyMuHPnDidPnuTRo0d5GuXJb+Hh4Vy7do133nknx/2s+vfvz5IlS9i0aROfffaZ1hqjrOnP\n4fn03rCwMK5evUrZsmX58ssvlXNxcXFMmTIFCwsLli5dirm5udZ1PvroIwIDAzlw4ABr165l6NCh\n+Xi3QgghhHiTGBRI2dnZ0bBhQyIjIwkMDNQaCdi0aROpqak4Oztr7SGlSWWtGWV4WXXq1EGlUvH3\n33/z559/am3Iq/mkumPHjrKHVAHSZF6D5wv4y5UrR8eOHRk6dKjBG9O6uLiwbds2li5dysGDB3ny\n5Ak2NjYMGjSIyZMnZwsEqlWrxu7du1m8eDHBwcFERERgbW1N+/btmTBhQrZMdMWKFWPjxo14eXmx\nc+dOVq1aReXKlZk4cSIffPBBngIpgGnTptGkSRNWr17NgQMHSE5Oply5cjRs2FAr/TbA999/j42N\nDdu2bWPdunWoVCqGDRvGmDFj2Llzp8HXnD17NkZGRhw5coTg4GAyMzPx8PCgRYsWwPNkDpaWlvz2\n229s2LABGxsbevbsyZQpU146w2Dr1q0JCgpi2bJlHDx4kOPHj2NmZoatrS0tW7aka9euL9VuftGM\nMPXv3z/HclWqVKF169YcOnSI/fv3a/2NunDhAhcuXFD+bWFhQeXKlfnkk08YO3as1mtv2rRpxMTE\n4OnpyVtvvaXzWosWLaJDhw7MmzePNm3aULNmzVe5RSGEEEK8oYzUBs5PuXfvHl9//TUJCQk0bdqU\nypUrc/nyZc6fP0/FihWZO3euVpYzzaaZmzZt0mrn4sWLygafT58+5dixY5QqVUrrjfG4ceO06ly+\nfJnZs2eTnp5OixYtKF++POfOnePq1au89dZbzJw5U2eabF0MWduQnJysN0vgi/4LySZeJ/K8C588\n88L1qs87L3+//qsyRvTQ+rftH+GyMLwQyUL8wifPvHDJ8y58RZVswuCsfXZ2dixYsIBNmzZx+vRp\nTp06RZkyZejSpQt9+vQxaGNWeB6QhYaGah1LSEjQOvZiIFWrVi3l2mfOnCElJQUbGxv69OmDq6ur\nwUGUEEIIIYQQQuQHg0ek/i1kROrNJs+78MkzL1wyIlXwZESqaMmn9YVPnnnhkudd+F77ESkhhBDi\nTfRi4CSEEELkB4PSnwshhBBCCCGE+H8SSAkhhBBCCCFEHkkgJYQQQgghhBB5JIGUEK+pVatW0a5d\nO2rWrIlKpWLFihVF3SWhh5eXFyqVivDw8KLuihBCCCEKiSSbEHoFBARw9OhRzp8/z4ULF0hMTMTN\nzY2lS5fqLH/79m1++OEHzp49S3R0NAkJCZQpU4Zq1arx4Ycf4ubmlqdU9ampqfj5+bF582Zu3LhB\namoqKpWK1q1bM2rUKCpXrqxV/sSJEwQGBhIeHs6tW7dITEzE1taWVq1aMW7cOKpXr57tGpGRkcye\nPZtz585RtmxZ3NzcmDRpEubm5lrl1Go1vXv3Ji0tje3bt2NiYmLwfbyM7du3M3PmTOrVq8fw4cMx\nNzencePGBXrNf6Nnz54RFBREUFAQp06d4vbt26Snp2Nvb0/nzp0ZM2aMQVs3bNmyhYkTJwLPN2Ae\nMGBAQXddCCGEEK85CaSEXosXL+bChQtYWVlRsWJFrly5kmP5Gzdu8Mcff/Duu+/i4uJC6dKlefTo\nESEhIXh4ePD777/j5+eHqWnuL7v09HT69evHiRMncHBwwNXVFXNzc86cOcPq1av5/fff2b59O7Vr\n11bqjBgxgocPH9K0aVN69eqFqakpf/31F35+fmzfvh0/Pz+aNm2qlL979y7u7u6UKlWKAQMGcPHi\nRXx9fXn69Clff/21Vn/Wrl3LqVOnCAwMLPAgCiAoKAiAdevWYWdnV+DX+7e6ceMGw4cPp3jx4rRs\n2RInJyeSkpIIDQ3F19eXHTt2sH37dsqWLau3jdu3b/PVV19hZWVFUlJSIfZeCCGEEK8zCaSEXrNm\nzaJixYpUr16diIgI+vbtm2P5pk2bcuHCBYyNtWeMPnv2jP79+xMeHs7u3bvp0SP3VMR79uzhxIkT\ntGrVCj8/P6VNU1NTFi5ciI+PD8uWLcPb21upM2LECHr37p0t8FiyZAnffvst06dP58CBA8rxrVu3\nkpyczP79+6latSoAffv2Zd26dXz11VcYGRkBcOvWLRYsWMCkSZO0AreCdO/ePYB/VRCVlJTEjRs3\nqFOnTqFd08rKinnz5uHu7q61t1JaWhrDhw/nwIEDeHt7M3fuXJ311Wo1Hh4elClThg8++IBly5YV\nVteFEEII8ZqTNVJCr/fff58aNWooAUVuzM3NswVRAGZmZnTu3BmA69evG9TWzZs3AXBycsrWpouL\nCwAPHz7UOj5u3Didgce4ceOwsLDg4sWLxMXFKcejo6MpV66cEkQBNGrUiJSUFK1yU6dOxd7envHj\nxxvUd11SU1P54YcfcHJyombNmrz11lv06tWLHTt2aJV7ca2NSqVSvgxx9epVRowYQZ06dXBwcKBH\njx4EBQXh7++PSqXC399fq7yjoyOOjo48efKEWbNm4ejoSLVq1fDy8lL6Y2trS3h4OJs2baJTp07U\nrFmTBg0a4OHhwYMHDwx+BnFxcXTs2BEnJyd+/PFHbt++bXDdl1WxYkWGDBmSbYNac3NzJkyYAEBE\nRITe+qtWrSIsLAxvb2+DN7kNCAiga9eu1KxZk7p16zJmzBju3r2brVyfPn1QqVSkpaXh4+ND69at\nqV69ujKFUAghhBCvNxmREgUuIyOD4OBgAN555x2D6mhGfkJCQhg+fLhWMKWZ9ta6dWuD2jIyMlKm\nE2adlqdSqXj48CG3b99WApXIyEgsLS2VqV6//fYbERER7Nq1y6ApibqkpaXx0UcfERERgYODA0OG\nDCElJYVdu3YxZswYzp8/z+effw7Ae++9h4eHB5s2bSI6OhoPDw+Dr3PlyhV69uxJfHw8Tk5OvPPO\nO9y8eZPhw4fToUMHvfWePXuGu7s78fHxtG3bFmtra6pUqaJVZsWKFRw6dIju3bvTvn17jh8/jr+/\nPxEREQQEBFCuXLlc+1euXDlGjhzJjh07mD9/PgsWLMDR0ZFevXrRrVs3SpcubfC95gfNej19UzUv\nX77MggUL+OSTT2jRogVhYWG5trlu3Tr2799Px44dadGiBadOnWLHjh1cuHCBffv2UaxYsWx1RowY\nQWRkJO3bt6dz585UqFDh1W5MCCGEEIVCAimR7+Li4lizZg1qtZqHDx9y6NAhoqKi6NWrF506dTKo\nDWdnZ7p06cLu3btxcnKidevWmJmZcfbsWY4fP86wYcMYMmSIQW3t3LmTxMREGjduTKlSpZTjbm5u\nLFmyhN69e9OlSxcuXrxIWFgYI0eOxMjIiLt37zJnzhzGjh1LvXr1XuZRALB8+XIiIiLo0KEDa9as\nUQIyDw8Punbtyg8//ICzszPNmjWjZcuWtGzZkoiICKKjo5kyZYrB1/niiy+Ij49n/vz5DB48WDke\nHBzMoEGD9Na7f/8+tWrVYsuWLXpHXUJCQggICNB6Dp6enqxcuZL58+crI1g5KV68OJ6ennz99ddE\nRESwbds2du/ezfTp0/n6669p3749bm5uODs7Y2FhYfB9v6yNGzcC0L59+2zn0tPTmThxIpUqVWLG\njBkGt3nw4EF27dql9YHBuHHj2LZtG4GBgTqntd6+fZvg4GAleDc1NSU9PT2vtyOEEEKIQiZT+0S+\ni4uLw9vbGx8fH3755Rdu3LjB6NGj8fHxMbgNIyMjfv75Zzw8PLh69SqrVq1i2bJlhIWF4ejoiKur\nq0EjRDdv3uTrr7/G1NQUT09PrXOVKlVi48aNVKxYkV9//ZVr164xYcIE5Y3zjBkzsLOz49NPP+XC\nhQv06dMHe3t76taty8yZM0lLSzPoXjZu3IiRkRGenp5afS5fvjyTJ08GwM/Pz9BHo9Pt27cJCwvD\n3t4+W9DUoUOHXEfvZs6cmePUtd69e2cLJqdMmULJkiXZtm0bqampBvfV2NiY999/n++//55Tp06x\nevVqXFxcOHToEKNGjaJRo0Z4eHhw+PBhMjMzDW43L/bt28evv/5KxYoVGTNmTLbzPj4+nDt3Dh8f\nHywtLQ1ud9iwYdlGXTUZ/k6fPq2zztSpU3NMdiGEEEKI15OMSIl85+DgwO3bt8nIyODevXvs2bOH\nRYsWcfz4cX755RfKlCmTaxtPnz5l0qRJhISEMG/ePFxcXLC0tOTkyZN88cUX9O7dm+XLlyvrpXSJ\njY1l4MCBPHz4kHnz5mll7NNo3Lgxf/zxR7bjW7ZsITg4mO3bt5ORkcGgQYMoVaoUq1evJioqijlz\n5mBmZpYtu9+LEhMTiYqKws7ODgcHh2zn33//fQDOnTuX2yPJ0fnz5wFo0qSJznVqzZs35/Dhwzrr\nWlhY5JoAokWLFtmOlSxZkrp16xIREcHly5dfatTO3NwcFxcXXFxcSEpKYs+ePWzcuBF/f3/8/f3x\n8fHB3d09z+3m5MSJE4wbN47ixYuzYsWKbFMKT548ydKlSxk1apTO10xOGjZsmO1YpUqVAEhISNBZ\np1GjRnm6hhBCCCFeDzIiJQqMiYkJKpWK4cOH8+2333Ly5EkWLVpkUN0ff/yRgIAApk+fzqBBg6hQ\noQIlSpTAycmJn3/+mWfPnjFz5ky99WNjY3F3d+fq1avMnj3b4GmAADExMXh6ejJixAgaN27M1q1b\nuXfvHgsXLqRDhw4MGzYMNzc31qxZQ0pKSo5tPX78GABbW1ud5zXrYTTlXtaTJ08AsLGx0Xm+fPny\neuuWK1cu14Qi+trVHNdc/2U9e/aMiIgIDh48SGRkJAClS5fO96yFf/75J4MGDcLY2JgNGzbw7rvv\nap1PT09n0qRJ1KhRg6lTp+a5/ZIlS2Y7phmFzMjI0FlH1kQJIYQQbyYZkRKFQrMOJacMaVlpEkq0\nbNky27m6detSunRpoqOjiYuLyzYt6v79+/Tr148rV64wb968PAVR8HytUenSpZU30pcvXwagfv36\nSpn69euzceNGoqKickygoXljrS+7neZ4iRIl8tTHF2nqx8TE6DwfGxurt64hWRn1tas5/jL9z8zM\n5OjRo2zbto1du3YRHx+PhYUFzs7OuLm50b59+2wbI7+KY8eO8fHHHytBVJMmTbKVSUpK4tq1awDU\nqFFDZztTp05l6tSpfPLJJ8yePfuV+2VoVkwhhBBCvF4kkBKFQrMvkqGb2WrWH72Y4hyepxJPTEwE\nyPZG+86dO7i7uxMVFcXChQsZOHBgnvoZEBDAnj172LJlS7a1MampqcoxQ9cEWVtbY29vz40bN7h2\n7Vq2N+eaNOdZg7SXUbduXQD++usvMjMzs03vO378+Cu1f/To0Wz7iD1+/Jjz589jYWFBrVq1DG4r\nMjKSbdu2sWPHDu7du4exsTEtW7bEzc2NLl26vHJQqcuRI0cYMmQIxYoVY8OGDXqn05mbm9O/f3+d\n586ePcu5c+do3rw5NWvW1BmICSGEEOK/QwIpkW/Onj1LnTp1sgVLSUlJyjQ8JycnrXOPHz/mwYMH\nlChRQmv6W/Pmzbl48SJLly6lWbNmWmmjvb29SU9Pp1GjRlhbWyvHo6Oj6du3L9HR0Xh5edGvX788\n9f/Ro0d89dVXDB48GEdHR+W4JhX7/v37lWAiKCiIYsWKYW9vn2u7/fr149tvv2Xu3LmsWLFCeT5x\ncXH4+voC8OGHH+apry9SqVS89957REREsH79eq2sfSEhIXrXRxlqy5YtDB06VGsdlJeXF48fP6Zf\nv34603q/6OHDh/To0YOoqCjgefA4cuRIXF1d9U59zA+hoaEMGzYMS0tLNm7cmONaLktLS73TT728\nvDh37hx9+/ZVEkgUpGfPnnHjxg3MzMyoVq1agV9PCCGEEHkjgZTQa+/evezduxf4/ylcf/31l5Jp\nrmzZslrrlHx8fDhx4gRNmzZFpVJhaWnJnTt3CAkJISEhgaZNmyqboGrs2bMHDw8P+vbtqwQVABMn\nTmT//v0cOXKEtm3b0q5dOywsLPjzzz85deoUFhYWfPPNN1pt9enTh1u3btGgQQMlmHqRu7t7tj2S\nNGbOnImFhQVffPGF1vFevXrh5eXF559/zqlTp7hx4wbh4eGMGTPGoIxuo0ePJiQkhMDAQDp27EiH\nDh1ISUkhICCA2NhYxo4dS/PmzXNtJzfz58/H1dWVL774guDgYGUfqd27d+Pi4kJgYKDORBSGaN++\nPT179qR79+7Y2tpy/Phxjh8/TpUqVbI9L32Sk5PJzMxk4sSJ9O7dW2fyjfx25coVhg0bxtOnT+nQ\noQOBgYEEBgZmK5eXNPOFJTo6mrZt22Jvb2/QHlZCCCGEKFwSSL2EjBH/vxeM7uXjRctkxY58aef8\n+fNs3rxZ69iNGze4ceMGAJUrV9YKpAYMGICVlRWnTp0iIiKClJQUSpUqRf369enevTsffvihwZva\nVqxYkcDAQH788UcOHDjApk2byMzMxNbWFnd3d8aNG5ftjfitW7cAOHPmDGfOnNHZ7nvvvaczkAoK\nCmLr1q34+flhZWWldc7S0pINGzYwc+ZM/P39KV68OMOHD2fatGkG3Yu5uTl+fn78/PPPbNu2jTVr\n1mBiYkKdOnX45ptvcHV1Naid3NSuXZsdO3awcOFCwsLCCAsL45133mHlypVcuXKFwMDAl542N2LE\nCDp37szKlSvZuXMnxYsXx93dnRkzZuSYyCIrlUpl8Bq5/PLgwQOePn0KwO7du9m9e7fOcq9jICWE\nEEKI15uRWq1WF3UnCtOdO3dyLZOcnJzjnjpZA6nXUX4FUq8j2az05YwfP54//viD0NDQPI0EeXl5\n4e3tzebNm3Um/hD571Vf47n9/fovyu1vtu0f4TkmZBH5q3z58vK8C5k888Ilz7vw5fcz12xdkhtJ\nfy7Ev0RmZqbO7ICHDx9mx44d1K5du1Cm0wkhhBBC/BfI1D4h/iXS0tJo1qwZLVu2xMHBARMTE/75\n5x8OHTqEubk58+bNK+ouCiGEEEL8a0ggJcS/hJmZGYMGDSIsLIxTp06RkpJC2bJl6datG+PHj88x\nW50QQgghhMgbCaSE+JcwMTFh7ty5+drmlClTmD59uqxLE0IIIYR4gayREkIIIYQQQog8kkBKCCGE\nEEIIIfJIAikhXlOrVq2iXbt21KxZE5VKxYoVK4q6SyIHkydPRqVSKfuZCSGEEOLfTdZICZ38/f3x\n8PDIsYyxsXG2N42pqan4+fmxefNmbty4QWpqKpUqVaJNmzaMGjWKypUrG3T9Z8+esW7dOs6fP8+5\nc+e4fPkyz549w8vLiw8//FBvvdjYWJYtW0ZwcDDR0dGYmZlRpUoVevbsyaBBg7C2ttYqHxkZyezZ\nszl37hxly5bFzc2NSZMmYW5urlVOrVbTu3dv0tLS2L59OyYmJgbdx8vavn07M2fOpF69egwfPhxz\nc3MaN25coNf8t0pOTmbv3r0cOHCAs2fPcufOHYyNjalZsyY9e/Zk2LBh2X7euvj6+vL9998D4Ofn\nR5s2bQq660IIIYR4jUkgJXSqW7eu3kDq2LFjhIWF0b59e63j6enp9OvXjxMnTuDg4ICrqyvm5uZE\nRkayevVqfv/9d7Zv307t2rVzvX5ycjKenp4A2NjYYGNjk+tmyrdu3aJbt27Exsby3nvv0b59e1JT\nU3BvxzUAACAASURBVAkNDWXu3Lls2bKFnTt3YmlpCcDdu3dxd3enVKlSDBgwgIsXL+Lr68vTp0/5\n+uuvtdpeu3Ytp06dIjAwsMCDKICgoCAA1q1bh52dXYFf79/s2LFjTJgwgdKlS9OyZUs6d+5MfHw8\n+/fvZ86cOezZswd/f38sLCz0tnH27Fl8fX2xsrIiKSmpEHsvhBBCiNeVBFJCp3r16ulNl929e3cA\nBg4cqHV8z549nDhxglatWuHn54ex8f/PHF20aBE+Pj4sW7YMb2/vXK9vaWnJ+vXrqVu3Lra2tnh5\neeVa73//+x+xsbFMmTJFKwjMyMigf//+hIWFERAQQN++fQHYunUrycnJ7N+/n6pVqwLQt29f1q1b\nx1dffYWRkRHwPEBbsGABkyZNMigIzA/37t0D+E8HUeHh4fTt2xdvb2/69ev30u3Y2NiwdOlSunXr\npjXylJiYSJ8+ffjzzz9Zu3Yto0eP1ln/6dOnTJw4kYYNG1KtWjW2bNny0n0RQgghxL+HrJESefL3\n339z8uRJ7OzscHJy0jp38+ZNAJycnLSCKAAXFxcAHj58aNB1zM3N6dChA7a2tgb3TXP9Tp06aR03\nMTFR+pr1+tHR0ZQrV04JogAaNWpESkoKcXFxyrGpU6dib2/P+PHjDe7Li1JTU/nhhx9wcnKiZs2a\nvPXWW/Tq1YsdO3ZolfPy8kKlUhEeHg6ASqVSvgxx9epVRowYQZ06dXBwcKBHjx4EBQXh7++PSqXC\n399fq7yjoyOOjo48efKEWbNm4ejoSLVq1fDy8lL6Y2trS3h4OJs2baJTp07UrFmTBg0a4OHhwYMH\nD176mRSWevXq4ebmlm36nrW1NaNGjQIgIiJCb/0FCxZw8+ZNfHx8sr2u9Vm/fj1OTk7UqFGDhg0b\nMm3aNB4/fpytnK7nr1KplOcvhBBCiNeXjEiJPNmwYQMA/fv3zzbFTTNaExISwvDhw7XedGqmqrVu\n3brA+la7dm1CQkIICgrSGk3LzMwkJCQEY2NjWrVqpRxXqVQ8fPiQ27dvK4FKZGQklpaWlC1bFoDf\nfvuNiIgIdu3ahanpy/26pKWl8dFHHxEREYGDgwNDhgwhJSWFXbt2MWbMGM6fP8/nn38OwHvvvYeH\nhwebNm0iOjo613VqWV25coWePXsSHx+Pk5MT77zzDjdv3mT48OF06NBBb71nz57h7u5OfHw8bdu2\nxdramipVqmiVWbFiBYcOHaJ79+60b9+e48eP4+/vT0REBAEBAZQrV+6lnk1R0/xM9f1sjxw5wqpV\nq5g1axY1atQwqM25c+cSGhpKx44dadOmDeHh4WzYsIHr16+zefPmbOVffP4lS5bM9vyFEEII8fqR\nQOolmKz4/1EEU1PT/8xmpSkpKWzduhUTExP69++f7byzszNdunRh9+7dODk50bp1a8zMzDhz5gwn\nTpxg2LBhDBkypMD6N3bsWIKCgvj+++8JDw+nfv36PHv2jNDQUGJiYli0aJFWgOXm5saSJUvo3bs3\nXbp04eLFi4SFhTFy5EiMjIy4e/cuc+bMYezYsXqnORpi+fLlRERE0KFDB9asWaO8affw8KBr1678\n8MMPODs706xZM1q2bEnLli2JiIggOjqaKVOmGHydL774gvj4eObPn8/gwYOV48HBwQwaNEhvvfv3\n/4+9+4+vuX78/3/H2g82zDY2k5aQWA0hVrRKL4ZCDUnpJdEreRXTu8ivipU+L1SvVz/0ihBa5mfM\nIisz+TmZE5M2sbA2IWbYj/bj+4fvOS/Hzrbz3I8zP27Xy6VL9nw8no/n4zz2dJz7eTyfj+cJtWjR\nQitWrFDt2rVt1tm0aZOio6OtxmHq1KmaO3eu3n777Wt2BsU8QxcSElKs7Ny5cxo7dqzuueceDR8+\n3O429+zZo++++84SzvPz8zVw4EBt27ZNiYmJateunVX9K8f/RnpPAQDgWsalfbDb2rVrlZmZqZCQ\nEJuXmtWoUUP//e9/FR4erl9//VXz5s3TnDlztG3bNt1zzz3q169fuWd17OHt7a21a9cqNDRUW7du\n1Zw5czRv3jz9+uuv6tOnj9VslCQ1btxYX331lfz8/LR48WIdPnxY//znPzV+/HhJ0vjx4+Xr66ux\nY8fqwIEDCgsLU0BAgNq0aaMpU6YoLy/Prn599dVXqlGjhqZOnWr1+r29vTVmzBhJl1aBq4i0tDRt\n3bpVAQEBxULTgw8+WOZM4JQpU0oMUZL0+OOPFwuT48aNU926dbV69Wrl5uaWv/PVZP78+dq0aZPa\ntGljcyXISZMm6ezZs5o9e7blfjl7jB071urvh5OTk+Uer71799rcp6zxBwAAVx9mpGA382V9Vy4y\nYZaTk6OXX35ZmzZtUkREhHr06CE3NzclJCRo8uTJevzxx/Xpp59a7peqbMeOHdOwYcOUnZ2tRYsW\nqWPHjsrOztaGDRv01ltv6dtvv9WaNWus7olq3769Vq1aVaytFStW6Pvvv9fXX3+tgoICPf3006pX\nr54+//xzpaamatq0abrpppuKre53pfPnzys1NVW+vr5q3rx5sfJ7771XkrR///4KvfakpCRJ0t13\n323zPp5OnTppy5YtNvd1dXVV69atS22/c+fOxbbVrVtXbdq00fbt25WSklKhWbuwsLAS71MKDw8v\ndoljly5dtHz5cklSZmam5s6dW2y/5557TvXq1bPZZkxMjKZOnaqGDRvqs88+00033WRVvm7dOq1Y\nsUIRERG65ZZbDL2Wu+66q9g2Pz8/SdLZs2eLldkz/gAA4OpDkIJdfvnlF+3evVt+fn7FFpkw++ij\njxQdHa233nrLalbEvGjE3/72N02ZMqXKgtTYsWP1888/a+PGjZYPph4eHnr66aeVm5urqVOnavbs\n2Xr//fdLbefkyZOaOnWqRowYofbt22vJkiXKyMjQJ598ok6dOkm6FFzmz5+vV155xbKcui3mBQZK\nWjSjYcOGVvXKKysrS9KlFeps8fb2LnFfLy+vMmdcSmrXvN18/PIaMGCAunTpYrXt2LFjWrZsmXr0\n6KE2bdpYlV3+PLJz587ZXNHRvLT9ldavX69Ro0bJ29tbUVFRxYLSmTNnNH78eN13331Wl0jay9Yx\nzTORhYWFxcrsGX8AAHD1IUjBLqUtMmFmXlAiODi4WFmbNm1Uv359HT9+XH/++adlMYfKcv78eW3f\nvl3169e3+e2+uU/79u0rs63XX39d9evX1//93/9JklJSUiRJd955p6XOnXfeqa+++kqpqam64447\nSmyrbt26klTi6nbm7R4eHmX2qzTm/U+ePGmz/NSpUyXua8+H+JLaNW+vaP9tLW++bds2S5Aqbfnz\nm2++WWlpaXYdZ+3atRo9erR8fHwUFRVlcwGJtLQ0/fnnn/rhhx9KfIC0+R7BN954QyNGjLDr2CUh\nRAEAcG0iSKFMOTk5WrFihWrVqmXzXhIz8z1DtpY4z83N1fnz5yWp2DLUlcF87PPnzysvL6/YMcx9\nuvISritFR0frm2++0YoVK4rNNOXm5lq22XtPkLu7uwICAvTbb7/p8OHDxT64m5c5vzyklYd5xubH\nH39UYWFhscv7du3aVaH2d+zYYXn+ltm5c+eUlJQkV1dXtWjRokLtO8LKlSs1ZswY+fr6atmyZSVe\nsufp6WlzMRXp0jgcOXLEMst6++23V2WXAQDAVYzFJlCm6OhonT17Vg888ECpzzMyX/b2n//8p1jQ\nmD17tvLz89W2bVu5u7tbtp87d06HDh3SiRMnKtTHBg0aqEWLFsrPzy926V5OTo4++OADSSq24MTl\nzpw5o0mTJumZZ57RPffcY9luXtZ948aNlm2xsbFycXFRQEBAmX0bNGiQioqKNH36dBUUFFi2//nn\nn5a+lhZQ7eHv768uXbooNTVVixYtsirbtGlTifdH2WvFihXF7uOaNWuWzp07p759+8rFxaVC7Ve1\nqKgovfzyy/L399fKlStLve/J399fM2fOtPlfhw4dJEkjRozQzJkz1a1btyrtd3Z2tg4dOmT3jBsA\nAHAcZqRQJvNlfUOGDCm13ksvvaSNGzfqhx9+0P3336+QkBC5urpq9+7dSkxMlKurq958802rfb75\n5huFh4drwIABxQLQhx9+qEOHDkn632IKX331lXbs2CHpUnB78sknLfXfeustPfPMM/rggw8UHx+v\nDh06KCcnR5s2bdLx48cVEBCgUaNGldj/KVOmyNXVVa+//rrV9v79+2vWrFmaMGGCEhMT9dtvv2nb\ntm164YUXSr0/yuwf//iHNm3apA0bNujhhx/Wgw8+qOzsbEVHR+vUqVMaNWqUJYRWxNtvv61+/frp\n9ddf1/fff295jlRMTIx69OihDRs22P1A2Ss98MAD6tu3rx555BE1atRIu3bt0q5du3TzzTcXG6+r\nzdatWzVu3DgVFhYqODi42EOJpUuXYFb0Er2qkJCQoMGDB+u+++6z2W8AAFB9CFIoVUpKinbt2lXq\nIhNmfn5+2rBhgz766CN99913ioqKUmFhoRo2bKiBAwfqxRdftLlyXUni4uKKreSWkJCghIQEy8+X\nB6lu3bpp3bp1+uSTT7Rjxw4tWLBAtWrVUtOmTTV69GiNGjWqxFXcYmNjtXLlSkVGRqpOnTpWZW5u\nblqyZImmTJmipUuXqnbt2nruuef06quv2vU6nJ2dFRkZqf/+979avXq15s+fr1q1aql169Z68803\n1a9fP3uHpFQtW7bUmjVrNGPGDG3dulVbt27VHXfcoblz5+rQoUPasGFDue9lGjFihHr27Km5c+dq\n7dq1ql27tgYOHKjx48eXupDF1eD48eOWRR6++uorm3WaNGlyVQYpAABw9apRVFRUVN2dcKTff/+9\nzDoXL160+5kuPDzTsRjv8hk9erRWrVqlzZs3Gwqzs2bN0uzZs7Vs2TKbi4ig8lX0HDfy/nWjKBjx\naKnljVZtK3VBFlQub29vxtvBGHPHYrwdr7LHvHHjxnbV4x4p4DpRWFhoc3XALVu2aM2aNWrZsqWh\nEAUAAICScWkfcJ3Iy8tTx44dFRwcrObNm6tWrVpKTk5WfHy8nJ2dFRERUd1dBAAAuG4QpIDrxE03\n3aSnn35aW7duVWJiorKzs9WgQQP16dNHo0ePVmBgYHV3EQAA4LphKEidPn1aS5culclkUlZWljw9\nPdWxY0eFhYVZLWldlvPnz2v58uVKSEjQmTNn5OHhoaCgIA0aNEheXl4299mzZ49iYmJ0/Phxy7Gb\nNWumPn36WJanBm5ktWrV0vTp0yu1zXHjxum1117jvjQAAIAr2B2kMjIyNHnyZGVmZqpDhw7y9/fX\noUOHFBMTo71792ratGl2rQiWlZWlSZMmKT09XYGBgQoODlZaWpri4uKUmJio6dOnq1GjRlb7LF68\nWGvWrJGHh4c6duwoDw8PZWRkKCEhQTt37tSLL75Y5c9zAQAAAAAzu4PUvHnzlJmZqWHDhik0NNSy\nfeHChVq3bp0iIyM1cuTIMtuJjIxUenq6+vTpo6FDh1q2x8TEaMGCBZo7d64mTpxo2X727FmtXbtW\n9erV08yZM62Wr96/f7/eeustRUVFEaQAAAAAOIxdq/ZlZGTIZDLJx8dHPXr0sCobOHCgXFxctGXL\nFuXk5JTaTk5OjuLj4+Xi4qIBAwZYlfXs2VM+Pj4ymUw6ceKEZfvJkydVVFSkFi1aFHsGUGBgoNzc\n3HTu3Dl7XgZQbQ4fPqzhw4erbdu28vf31x133FHdXUIpjh07Jn9/f40ZM6a6uwIAAK5SdgWppKQk\nSVJQUJBq1rTexc3NTa1atVJubq5SUlJKbSc5OVl5eXlq1aqV3NzcrDtSs6aCgoKsjiddesirk5OT\nDh06VCwwHThwQNnZ2brzzjvteRkoh9jYWA0ePFh33323brvtNnXp0kUjR47U7t27Dbf1+++/Kzw8\nXO3bt9ett96qe+65R1OmTNHZs2dL3Cc5OVnPP/+87rrrLjVr1kzBwcGaOXOmsrOzi9XNy8vTu+++\nq86dO6tVq1YKCwvTvn37bLYbHx8vf39/bdy40fDrMKqgoEDDhw/X999/r+7duys8PFwvvvhilR/3\nerV//37NmjVLffv2Vbt27RQQEKC7775bo0aNKvH3faU///xT7dq1k7+/f6U9EBkAANxY7Lq0z/wQ\nWz8/P5vlvr6+MplMSk9PLzXU2NPO5fUkyd3dXUOGDNEXX3yh8PBwq3ukfvzxR9111112XVII4yIi\nIvTxxx/L09NTPXv2VIMGDXTkyBF9++23iomJ0QcffKDHH3/crrZSU1PVt29fnTp1Sj169FDz5s2V\nmJioefPmKS4uTqtXr1aDBg2s9tmzZ48GDhyo/Px89e7dW40bN9a2bdv03nvv6YcfftDSpUvl4uJi\nqf/222/rs88+U69eveTn56cVK1ZowIAB2rx5s9V9dxcuXNCrr76qxx57TA8//HDlDFYpjh49quTk\nZA0ZMkT/7//9vyo/3vVu/PjxSkxM1F133aXQ0FDVqVNHSUlJ+vrrr7Vu3Tp98skn6tWrV6ltvPba\na7pw4YKDegwAAK5HdgWpixcvSpJq165ts9y8vawPJva2Y65n1rt3b/n4+OiTTz7Rd999Z9nu6+ur\nkJCQYpf8XS42NlaxsbGSpBkzZsjb27vUPkrSiRMn5ORk/4KGRupeK/744w/NmTNHPj4+2rRpk3x8\nfCxlP/zwgx5//HHNnDlTgwYNsqu9iRMn6tSpU4qIiNBzzz1n2T5lyhR9+umn+te//qV//etflu0F\nBQUaN26csrOztXDhQvXs2VPSpYfOjhgxQtHR0Zo3b55eeuklSVJRUZEWL16sJ554Qh988IEkqU+f\nPurfv79WrVql0aNHW9qeMWOGcnJyFBER4ZDfnflJ2+bZ1WtRRftdUFCgxMREdejQocJ9CQsL0yef\nfKJbb73Vavvy5cv14osv6rXXXlPPnj3l7Oxsc/+oqCjFxMTo3Xff1WuvvaYaNWoUe321atWSdGmm\nvDp+ZxU5pouLi13vczeSE2WUOzk5MWYOxHg7HmPuWIy341XXmF8Tn+q+/vprRUZGKjQ0VD179lT9\n+vWVlpamyMhI/fvf/1Zqaqqeeuopm/t2795d3bt3t/xs/lBbmtzcXMsHqbI4OTldl0tDp6amqrCw\nUO3atZOnp6fVa+zcubPc3d11+vRpu157amqq4uLidPPNN2vo0KFW+4SHh2vRokVatmyZJk+ebAnT\nP/zwg5KTk9W5c2d1797dso+Tk5Nef/11RUdHa+HChXrhhRdUo0YNnTp1StnZ2QoKCrLUNc+OHj16\n1LJt586dmj9/vubMmaO6deuW+3f3008/6T//+Y927typrKws+fj46KGHHtKYMWOsZr/8/f0tf545\nc6Zmzpxped3jxo0r9RhFRUWaN2+eFi9erKNHj1pmBl977TXLTNrOnTst9ZcuXarw8HDNnj1bDRs2\n1EcffaT9+/crKytLaWlplv506dJFH374oSIiIrR582ZduHBBLVq00PPPP6/+/fsX60dFzvGffvpJ\nK1as0Jo1a+Ts7GzV3/L6+9//LknF+tSvXz/NnDlTR44c0f79+3XXXXcV2zctLU0TJ07U4MGDdf/9\n90u6NM5XtlVQUCDpUnA/cuSI3n77bW3ZskUXL17U7bffrvDw8GKzmfaMvz0q+p6Sm5tr1/sc/ic/\nP58xcyBvb2/G28EYc8divB2vsse8cePGdtWzK0iVNFNkZt5ep06dSmnn8hmrpKQkLVmyRJ06ddIz\nzzxj2d6sWTO98sorevnll7V27Vo9/PDDxZZNR/ndeuutcnZ21t69e/Xnn39aXXa3Y8cOnT9/3jJL\nVJZt27ZJkrp161bsHjt3d3d17NhRmzdv1o8//qiuXbtKkrZu3SpJCgkJKdbeLbfcombNmunw4cP6\n7bffFBAQoAYNGsjV1dXqHpmffvpJktSkSRNJUnZ2tsaNG6fQ0FD16dPHzpEobuPGjRo5cqSKiorU\nu3dvNWnSRD/99JO++OILffvtt1q1apWaNm0q6VJgOnbsmJYtW6YuXbqoS5cukmT5f2lef/11ffHF\nF/L19dWQIUN000036dtvv9XevXuVn59f4qzFunXrFBcXpwceeEBPP/20jh8/blV+9uxZ9e3bV3Xr\n1tWgQYOUmZmp6OhojR49WhkZGXrhhRfKPTbSpeC8atUqrVq1Sr/++qskqXXr1nryyScr1K49brrp\nJkmy+UVIUVGRxowZIw8PD02dOrXUe/PMjh8/rt69e6tp06Z6/PHHLauIPvvss/rqq6907733Ftun\nrPEHAADXB7uClDmVpaen2yzPyMiQVPK9T0bbuTwF/vjjj5KkNm3aFKvv4uKi5s2ba9euXTpy5AhB\nqhJ5enrq9ddf15tvvqmQkBD17NlTnp6eSk1N1caNG9WtWze9++67drVl/jDdrFkzm+W33nqrNm/e\nrMOHD1uClD37HD58WIcPH1ZAQIBq1qypIUOG6PPPP9e5c+fk6+urFStWyMPDwzLL8q9//UtnzpzR\n22+/bWgsLnfhwgWNGTNG+fn5Wr58ue655x5L2UcffaS3335br732miIjIyVdeqDttm3bLEGqrFko\ns507d+qLL75Qs2bNFB0dbbl8dfz48XriiSeUkZFhCYhX+v7777Vo0SI98MADNst//vln9enTR598\n8okl2I4ePVqhoaF699131atXL91yyy12j4l06WHda9as0cqVK7Vnzx5JUkBAgF5++WX1799fLVq0\nMNReefz4449KTk6Wr6+vWrVqVaz8s88+0/bt2/Xll1/Kw8PDriC1fft2jRs3TuHh4ZZt/fv315Ah\nQzRnzhybQaqs8QcAANcHu4KUOcSYTCYVFhZazSpkZ2fr4MGDcnFxKfPDUsuWLeXs7KyDBw8qOzvb\nauW+wsJCmUwmq+NJ/7t8p6Qlzs3br9V7T65mI0aM0M0336xx48ZpyZIllu0BAQEaOHCg3deiZmVl\nSZLq1q1rs9z8IOfLf8fmfUp6yLO5rcv3mThxotzc3LR69WqdPXtWgYGBmjx5svz8/JSYmKi5c+fq\nvffek5eXl2bNmqXFixfr9OnTat26taZNm6aOHTuW+Vo2bNigs2fPql+/flYhSpKef/55LVq0SPHx\n8UpLS7O6rM+oZcuWSZJeeuklq3sAnZ2dNWHChFJXmuvRo0epH+Jr1aqliRMnWv09btq0qZ599lnN\nnj1bK1assAoOJbl48aI2bNiglStXKj4+Xvn5+WrYsKGGDx+ufv36qX379va81Epx5swZvfzyy5Kk\nN954o9iMVHJyst599109/fTThp4516RJE0u7ZiEhIfL391diYqLNfcoafzhGwYhHq7sLAIDrnF3L\nn/v6+iooKEgnT57Uhg0brMqioqKUm5urrl27ytXV1bI9LS2t2H0Brq6u6tatm3Jzcy0fFM3Wr1+v\nkydPKigoyGpmyfzNcmxsrP7880+rfRITE/XLL7/opptu0u23327PS4EBH3/8sUaOHKkBAwZo27Zt\nOnTokNavX69bbrlFo0eP1vTp06u7i1ZcXFw0YcIE7dy5U7/88otWrFihtm3bKi8vT+Hh4QoJCdHj\njz+uefPmafbs2Xrqqae0ePFiy8qQJ0+eLPMY5ksHbc1EODk5WcLV/v37K/RazPt36tSpWFn79u1L\n/eKgbdu2pbbt7+9vufTwcubLDe3t+8SJEzV69Ght3rxZ/fr1U2RkpHbv3q233nqrQiFq/fr1mjVr\nltV/69evL7H+xYsX9eyzz+rIkSMaNWqUHnnkEavyv/76Sy+99JIaNmyoSZMmGepLmzZtbF4m2Lhx\nY2VmZtrcp6zxBwAA1we7p3GGDx+uyZMna/78+dq3b5+aNGmilJQUJSUlyc/PT4MHD7aqP3bsWEmX\ngtblBg8erKSkJEVHRys1NVXNmzfX8ePHtXv3btWrV0/Dhw+3qt+5c2fdeeed2rdvn8aOHauOHTta\nFpvYs2ePioqKNGTIkBJnLlA+27ZtU0REhEJDQ/XGG29Ytt95552aN2+e7rvvPn366ad6+umny7wM\nzNaM0+VszViZ9zGXXcncVkmzXJd77733lJGRoS+//FKSNGfOHN13332Wy+xuu+023XPPPVqwYIH+\n7//+r9S2zP1p2LChzXLzlwAlfci2l/k4l6+WaFarVi15enqWuG9JfTMraSbRvJ+9D7i+7bbb5Ozs\nrLy8PP3www9q0KCB6tWrZ3keXHmtX7++2BctAwYMsHlP3sWLFzV06FDt2rVLI0eO1MSJE4vV+fDD\nD7V//34tW7aszPs4r1TS+eXk5KTCwkKbZWWNPwAAuD7YHaR8fX31zjvvKCoqSnv37lViYqI8PT3V\nq1cvhYWFyd3d3a52PDw8FBERoWXLlikhIUE///yzPDw8FBISokGDBsnLy8uqfs2aNTVhwgRt2LBB\n27ZtU0JCgnJzc+Xu7q527dopNDS0wh/cUJx5yfjg4OBiZW5ubmrXrp2++eYb7d+/v8wgddttt0mS\nDh8+bLP8yJEjkqzvhyrPPrbs379fH3/8sd555x35+fkpKytLGRkZVqvT+fv7q0GDBkpOTi61Lel/\nAa+k2asTJy4ttGxPwCuN+e/TyZMni41vQUGBzpw5Y3numlElrWrzxx9/SLK/76NHj9ZTTz2l6Oho\nrVq1Sp999pn++9//6tZbb1W/fv3Ut2/fct0b9f777+v9998vs9758+c1dOhQ7dy5U6NGjbIZoqRL\ns4hFRUUKCwuzWZ6QkCB/f3/VrVtXP//8s+H+AgCAG5OhG4u8vb01atQou+peORN1OXd3dw0bNkzD\nhg2zqy0nJyf17t1bvXv3tqs+Ki4vL0/SpUUEbDFvN6+SVhpzGIuPjy92j9358+eVkJAgNzc33X33\n3Zbt9957r/79738rLi5O//znP63a++2333T48GE1adKk1BCXn5+v8PBwdenSpdiKcebXZ5abwbdo\nigAAIABJREFUm1vm65CkwMBASZdm7K6chc3Pz9euXbskqdQHU9t7nP3792vXrl3FXuOePXsqtDx2\nWlqajh07pptvvtlq+/bt2y3Htlf9+vX11FNP6amnnlJaWppltb733ntP7733ntq0aaN+/frp0Ucf\nLXFxjPI4d+6chgwZoj179uill17Sa6+9VmLdbt26FXvYs3Rp4ZA1a9bIx8dH3bt3t7pnEwAAoCx2\n3SOFG4/53pwlS5YUW2Xx+++/V0JCglxdXa0esPrXX3/p0KFDSk1NtaofEBCg+++/X8eOHdOCBQus\nymbOnKmLFy/q8ccft1r2vkuXLmrRooV27Nihb7/91rK9sLBQERERkqSnn35aNWrUKPE1fPjhh0pN\nTbV60K+Hh4d8fX21adMmSxjZvn27zp8/r5YtW5Y5LubnmH399deWFSXN5s6dq6NHj6pr164VWmhC\nkmX25N///rfVpXZ5eXmaMWNGhdouKChQRESE1aVpR48e1eeffy4nJyc99thj5WrX399fo0eP1nff\nfaeNGzdq1KhROnPmjCIiItS5c2e7vzgpy9mzZ/XEE09oz549euWVV0oNUdKl506Zn+F1+X+vv/66\npEvn58yZMzVt2rRK6V9p0tLSdOjQIWVnZ1f5sQAAQNViqTvY1KdPH3355ZfasmWLZfnzhg0bKiUl\nRbGxsSoqKtKECROsvunPyMjQ/fffryZNmhR78Orbb7+tvn37avLkyfrhhx/UokUL7dmzR9u2bVOz\nZs2KfRiuVauWZs+erYEDB2rkyJHq3bu3/P39tXXrVu3du1cdO3bUiBEjSux/cnKyPvjgA02ZMqXY\nzMuoUaM0ZcoU9e/fX+3atdOqVatUp04dy4NeS1OnTh3Nnj1bzz//vMLCwtSnTx81btxY+/bt0+bN\nm9WwYUO7l4UvTZcuXTRkyBAtWbJEDzzwgHr37i0nJydt3LhRdevWla+vb7FnctnrjjvuUGJionr2\n7Kn777/f8hypzMxMTZo0SQEBARXuf+vWrdW6dWu9/vrr2r59u1atWmVZlbOinnvuOZlMJgUEBKiw\nsFCzZs0qVqdHjx6GZtYc5cUXX1RCQoJWrVplcyERAABw7SBIwaaaNWtq0aJFWrBggb7++mutX79e\n2dnZql+/vh588EENHz5c999/v93tBQQEKCYmRjNnzlRcXJy+//57y1LZ4eHhql+/frF92rdvb9ln\n8+bNunDhgpo0aaKxY8fqxRdflIuLi81jFRQUKDw8XO3atbMZjp599lllZWVp0aJF2r9/v1q3bq03\n3njD5sIOtvTo0UOrV6/Wf/7zH8XFxSkrK0s+Pj56+umnNWbMmHLfu3SlGTNmqHnz5lq8eLEWLVok\nT09P9ezZU+PHj1eHDh0MP+vJrH79+lq8eLEiIiK0dOlSnT9/Xi1atFBERITVvWOVoUaNGgoODlZw\ncLAKCgoqpc1jx45JuvTg39mzZ9us06RJk6sySAEAgOtHjaKioqLq7oQj/f7772XWuXjxotVlZqVx\ncnKq0P0qMIbxluXBxX379tXHH39saF9/f3916dJFy5cvt3sfxtyxKjreRt6/rmdGniPVaNW2Ehdh\nQeXz9vZmvB2MMXcsxtvxKnvMGzdubFc97pECrlJ//PFHsSW2s7OzNXXqVEmyuRw4AAAAHINL+4Cr\n1Ny5c7V69Wp16dJFjRo10h9//KEffvhB6enpevDBB4s9eBYAAACOQ5ACrlJdu3bVgQMHFB8fr7Nn\nz6pWrVpq1qyZhg8frueee67UFQsBAABQtQhSwFWqa9eu6tq1a6W2mZaWVqntAQAA3Ki4RwoAAAAA\nDCJIAQAAAIBBBCkAAAAAMIggZcMN9mgtANcR3r8AAHAMglQJ+DAC4FrD+xYAAI5DkLKhVq1aKigo\nqO5uAIAhBQUFcnJiMVYAAByBIGWDi4uLcnJylJ+fzze8AK56RUVFys/PV05Ojpydnau7OwAA3BD4\n6tKGGjVqqE6dOsrLy1Nubq5lmy0uLi6WOqh6jLfjMeaOZXS8zV/2ODk5qU6dOjyoGQAAByFIlaBG\njRpycXGRi4tLqfW8vb116tQpB/UKjLfjMeaOxXgDAHBt4NI+AAAAADCIIAUAAAAABhGkAAAAAMAg\nghQAAAAAGESQAgAAAACDWLUPAID/X8GIR61+rvXZmmrqCQDgaseMFAAAAAAYRJACAAAAAIMIUgAA\nAABgEEEKAAAAAAwiSAEAAACAQQQpAAAAADCIIAUAAAAABhGkAAAAAMAgghQAAAAAGESQAgAAAACD\nCFIAAAAAYBBBCgAAAAAMIkgBAAAAgEFO1d0BAACuVgUjHrX6udZna6qpJwCAqw0zUgAAAABgEEEK\nAAAAAAwiSAEAAACAQQQpAAAAADCIIAUAAAAABhGkAAAAAMAgghQAAAAAGESQAgAAAACDDD2Q9/Tp\n01q6dKlMJpOysrLk6empjh07KiwsTO7u7na3c/78eS1fvlwJCQk6c+aMPDw8FBQUpEGDBsnLy6vE\n/fbt26f169crOTlZFy5ckIeHh5o2barQ0FC1b9/eyEsBAAAAgHKzO0hlZGRo8uTJyszMVIcOHeTv\n769Dhw4pJiZGe/fu1bRp0+Th4VFmO1lZWZo0aZLS09MVGBio4OBgpaWlKS4uTomJiZo+fboaNWpU\nbL/FixdrzZo18vLyUocOHeTh4aFz587pyJEjOnDgAEEKAAAAgMPYHaTmzZunzMxMDRs2TKGhoZbt\nCxcu1Lp16xQZGamRI0eW2U5kZKTS09PVp08fDR061LI9JiZGCxYs0Ny5czVx4kSrfWJjY7VmzRrd\nf//9ev755+XkZN3t/Px8e18GAAAAAFSYXfdIZWRkyGQyycfHRz169LAqGzhwoFxcXLRlyxbl5OSU\n2k5OTo7i4+Pl4uKiAQMGWJX17NlTPj4+MplMOnHihGX7X3/9pa+++kre3t42Q5Qkm9sAAAAAoKrY\nFaSSkpIkSUFBQapZ03oXNzc3tWrVSrm5uUpJSSm1neTkZOXl5alVq1Zyc3Oz7kjNmgoKCrI6niT9\n9NNPOnfunDp16qQaNWpoz549Wr16tWJiYpScnGxP9wEAAACgUtk1lfP7779Lkvz8/GyW+/r6ymQy\nKT09XXfeeWeF2rm8niT9+uuvkiRnZ2e9+uqrOnbsmNU+d9xxh8aNG6e6deva81IAAAAAoMLsClIX\nL16UJNWuXdtmuXn7hQsXKqUdcz1JyszMlCStWbNGTZo00VtvvaWAgAD98ccfWrRokUwmk2bPnq03\n3njDZpuxsbGKjY2VJM2YMUPe3t6l9tEoJyenSm8TJWO8HY8xdyzGu3KcKLuKxeVjXtZ+/G4qjnPc\n8Rhzx2K8Ha+6xvyqv7moqKhIklSrVi29+uqratiwoSSpadOmeuWVVzRmzBgdOHBAycnJatmyZbH9\nu3fvru7du1t+PnXqVKX2z9vbu9LbRMkYb8djzB2L8Xa8/Px8u8ec303FcY47HmPuWIy341X2mDdu\n3NiuenbdI2Vrpuhy5u116tSplHYun7Ey/zkgIMASosxcXFws91UdOnSo1GMDAAAAQGWxK0iZU1l6\nerrN8oyMDEkl3/tktJ3LU6D5zyWFNPP2vLy8Uo8NAAAAAJXFriDVpk0bSZLJZFJhYaFVWXZ2tg4e\nPCgXFxe1aNGi1HZatmwpZ2dnHTx4UNnZ2VZlhYWFMplMVseTpDvvvFM1atTQ8ePHix1bkmXxiStn\nqwAAAACgqtgVpHx9fRUUFKSTJ09qw4YNVmVRUVHKzc1V165d5erqatmelpamtLQ0q7qurq7q1q2b\ncnNztWzZMquy9evX6+TJkwoKClKjRo0s2318fHT33Xfr1KlTiomJsdrHZDLJZDKpTp06atu2rX2v\nGAAAAAAqyO7FJoYPH67Jkydr/vz52rdvn5o0aaKUlBQlJSXJz89PgwcPtqo/duxYSZeC1uUGDx6s\npKQkRUdHKzU1Vc2bN9fx48e1e/du1atXT8OHD7d57CNHjuiLL75QYmKiZdW+hIQE1axZU88//3yJ\nKwECAAAAQGWzO0j5+vrqnXfeUVRUlPbu3avExER5enqqV69eCgsLk7u7u13teHh4KCIiQsuWLVNC\nQoJ+/vlneXh4KCQkRIMGDZKXl1exfby8vPTuu+9q+fLl2r17tw4cOKDatWvr7rvvVv/+/dW8eXP7\nXzEAAAAAVJCh5c+9vb01atQou+peORN1OXd3dw0bNkzDhg2z+9h169bVs88+q2effdbufQAAAACg\nKth1jxQAAAAA4H8IUgAAAABgkKFL+wAAuN6c6B9c3V0AAFyDmJECAAAAAIMIUgAAAABgEEEKAAAA\nAAwiSAEAAACAQQQpAAAAADCIIAUAAAAABhGkAAAAAMAgghQAAAAAGESQAgAAAACDCFIAAAAAYBBB\nCgAAAAAMIkgBAAAAgEEEKQAAAAAwiCAFAAAAAAYRpAAAAADAIIIUAAAAABhEkAIAAAAAgwhSAAAA\nAGAQQQoAAAAADCJIAQAAAIBBBCkAAAAAMIggBQAAAAAGOVV3BwAAuFYUjHjU8udan62pxp4AAKob\nM1IAAAAAYBBBCgAAAAAMIkgBAAAAgEEEKQAAAAAwiCAFAAAAAAYRpAAAAADAIIIUAAAAABhEkAIA\nAAAAgwhSAAAAAGCQU3V3AACAiioY8Wh1dwEAcINhRgoAAAAADCJIAQAAAIBBBCkAAAAAMIggBQAA\nAAAGEaQAAAAAwCCCFAAAAAAYRJACAAAAAIMIUgAAAABgkKEH8p4+fVpLly6VyWRSVlaWPD091bFj\nR4WFhcnd3d3uds6fP6/ly5crISFBZ86ckYeHh4KCgjRo0CB5eXmVuX98fLw+/PBDSdLzzz+vhx56\nyMjLAAAAAIAKsTtIZWRkaPLkycrMzFSHDh3k7++vQ4cOKSYmRnv37tW0adPk4eFRZjtZWVmaNGmS\n0tPTFRgYqODgYKWlpSkuLk6JiYmaPn26GjVqVOL+p06d0ueffy5XV1fl5OTY230AAAAAqDR2B6l5\n8+YpMzNTw4YNU2hoqGX7woULtW7dOkVGRmrkyJFlthMZGan09HT16dNHQ4cOtWyPiYnRggULNHfu\nXE2cONHmvkVFRfrkk0/k4eGhTp06ae3atfZ2HwAAAAAqjV33SGVkZMhkMsnHx0c9evSwKhs4cKBc\nXFy0ZcuWMmeIcnJyFB8fLxcXFw0YMMCqrGfPnvLx8ZHJZNKJEyds7v/NN99o//79euGFF+Ti4mJP\n1wEAAACg0tkVpJKSkiRJQUFBqlnTehc3Nze1atVKubm5SklJKbWd5ORk5eXlqVWrVnJzc7PuSM2a\nCgoKsjre5Y4fP64lS5YoNDRUrVu3tqfbAAAAAFAl7ApSv//+uyTJz8/PZrmvr68kKT09vVLaMdcz\nKygo0Icffihvb289+eST9nQZAAAAAKqMXfdIXbx4UZJUu3Ztm+Xm7RcuXKiUdsz1zJYvX64jR45o\n2rRpcnZ2tqfLFrGxsYqNjZUkzZgxQ97e3ob2L4uTk1Olt4mSMd6Ox5g7FuNdPrYvCK9a/J7Kh3Pc\n8Rhzx2K8Ha+6xtzQ8ufVISUlRatWrdIjjzyili1bGt6/e/fu6t69u+XnU6dOVWb35O3tXeltomSM\nt+Mx5o7FeF87+D2VD+e44zHmjsV4O15lj3njxo3tqmfXpX0lzRSZmbfXqVOnUtox1zNf0ufn56dB\ngwbZ01UAAAAAqHJ2zUiZU1lJ90BlZGRIKvneJ6PtmOvl5ORY6g4ZMsTmPp9++qk+/fRT9erVS3//\n+99LPT4AAAAAVAa7glSbNm0kSSaTSYWFhVYr92VnZ+vgwYNycXFRixYtSm2nZcuWcnZ21sGDB5Wd\nnW21cl9hYaFMJpPV8W666SY9+OCDNts6cuSIjhw5olatWqlx48bluuwPAAAAAMrDriDl6+uroKAg\nmUwmbdiwweqBvFFRUcrNzVX37t3l6upq2Z6WliZJ8vf3t2xzdXVVt27dFBsbq2XLllk9kHf9+vU6\nefKkgoKC1KhRI0mSs7Oz/vGPf9jsU1RUlI4cOaL7779fDz30kIGXDAAAAAAVY/diE8OHD9fkyZM1\nf/587du3T02aNFFKSoqSkpLk5+enwYMHW9UfO3aspEuB53KDBw9WUlKSoqOjlZqaqubNm+v48ePa\nvXu36tWrp+HDh1fCywIAAACAqmN3kPL19dU777yjqKgo7d27V4mJifL09FSvXr0UFhYmd3d3u9rx\n8PBQRESEli1bpoSEBP3888/y8PBQSEiIBg0aJC8vr3K/GAAAAABwhBpFRUVF1d0JR7ryYb8VxRKX\njsV4Ox5j7liMd/kUjHjU4ces9dkahx/zesA57niMuWMx3o53VS9/DgAAAAD4H4IUAAAAABhEkAIA\nAAAAgwhSAAAAAGAQQQoAAAAADCJIAQAAAIBBBCkAAAAAMIggBQAAAAAGEaQAAAAAwCCCFAAAAAAY\nRJACAAAAAIMIUgAAAABgEEEKAAAAAAwiSAEAAACAQQQpAAAAADCIIAUAAAAABhGkAAAAAMAgghQA\nAAAAGESQAgAAAACDnKq7AwAAXIsKRjxq9XOtz9ZUU08AANWBGSkAAAAAMIggBQAAAAAGEaQAAAAA\nwCCCFAAAAAAYRJACAAAAAIMIUgAAAABgEEEKAAAAAAwiSAEAAACAQQQpAAAAADCIIAUAAAAABjlV\ndwcAADCqYMSj1d0FAMANjhkpAAAAADCIIAUAAAAABhGkAAAAAMAgghQAAAAAGESQAgAAAACDCFIA\nAAAAYBDLnwMAUAmuXJK91mdrqqknAABHYEYKAAAAAAwiSAEAAACAQQQpAAAAADCIIAUAAAAABhGk\nAAAAAMAgghQAAAAAGESQAgAAAACDDD1H6vTp01q6dKlMJpOysrLk6empjh07KiwsTO7u7na3c/78\neS1fvlwJCQk6c+aMPDw8FBQUpEGDBsnLy8uqblZWlnbt2qU9e/bo6NGj+vPPP+Xk5KSmTZvqgQce\nUEhIiGrWJA8CAAAAcBy7g1RGRoYmT56szMxMdejQQf7+/jp06JBiYmK0d+9eTZs2TR4eHmW2k5WV\npUmTJik9PV2BgYEKDg5WWlqa4uLilJiYqOnTp6tRo0aW+tu3b9fcuXPl6empNm3ayNvbW2fPntWu\nXbs0Z84cJSYmKjw8XDVq1CjfCAAAAACAQXYHqXnz5ikzM1PDhg1TaGioZfvChQu1bt06RUZGauTI\nkWW2ExkZqfT0dPXp00dDhw61bI+JidGCBQs0d+5cTZw40bK9cePGevXVV9W+fXurmacnn3xSEyZM\n0M6dO7Vz50517tzZ3pcCAAAAABVi1zVxGRkZMplM8vHxUY8ePazKBg4cKBcXF23ZskU5OTmltpOT\nk6P4+Hi5uLhowIABVmU9e/aUj4+PTCaTTpw4YdkeGBioDh06FLt8r379+nr44YclSQcOHLDnZQAA\nAABApbArSCUlJUmSgoKCigUaNzc3tWrVSrm5uUpJSSm1neTkZOXl5alVq1Zyc3Oz7kjNmgoKCrI6\nXlmcnJws+wIAAACAo9iVQH7//XdJkp+fn81yX19fSVJ6enqltGOuV5qCggJt3rxZktS2bdsy6wMA\nAABAZbHrHqmLFy9KkmrXrm2z3Lz9woULldKOuV5plixZomPHjqldu3alBqnY2FjFxsZKkmbMmCFv\nb+8y2zbCycmp0ttEyRhvx2PMHYvxts+JsqtUO36PtnGOOx5j7liMt+NV15gbWv78ahETE6Po6Gj5\n+/vrn//8Z6l1u3fvru7du1t+PnXqVKX2xdvbu9LbRMkYb8djzB2L8b5+8Hu0jXPc8Rhzx2K8Ha+y\nx7xx48Z21bMrSJU1U2TeXqdOnUppp6QZK0lav369FixYoCZNmmjKlCmGnl8FAAAAAJXBriBlTmUl\n3QOVkZEhqeR7n4y2U1IKXLdunRYuXKibb75ZU6ZMUb169cruPAAAAABUMrsWm2jTpo0kyWQyqbCw\n0KosOztbBw8elIuLi1q0aFFqOy1btpSzs7MOHjyo7Oxsq7LCwkKZTCar411u9erVWrhwoQICAjR1\n6lRCFAAAAIBqY1eQ8vX1VVBQkE6ePKkNGzZYlUVFRSk3N1ddu3aVq6urZXtaWprS0tKs6rq6uqpb\nt27Kzc3VsmXLrMrWr1+vkydPKigoSI0aNbIqW758ub788ks1a9ZMU6ZMUd26dQ29SAAAAACoTHYv\nNjF8+HBNnjxZ8+fP1759+9SkSROlpKQoKSlJfn5+Gjx4sFX9sWPHSroUtC43ePBgJSUlKTo6Wqmp\nqWrevLmOHz+u3bt3q169eho+fLhV/bi4OEVFRalmzZpq1aqVYmJiivWtYcOGCgkJsfelAAAAAECF\n2B2kfH199c477ygqKkp79+5VYmKiPD091atXL4WFhdm96IOHh4ciIiK0bNkyJSQk6Oeff5aHh4dC\nQkI0aNAgeXl5WdX/448/JF269M9WiJKk1q1bE6QAAAAAOEyNoqKiouruhCPZ87BfI1ji0rEYb8dj\nzB2L8bZPwYhHq7sLZar12Zrq7sJViXPc8Rhzx2K8Ha+6lj+36x4pAAAAAMD/EKQAAAAAwCC775EC\nAKC6XAuX8gEAbiwEKQAAqsCV4Y97pgDg+sKlfQAAAABgEEEKAAAAAAwiSAEAAACAQQQpAAAAADCI\nIAUAAAAABhGkAAAAAMAgghQAAAAAGESQAgAAAACDCFIAAAAAYJBTdXcAAABbCkY8Wt1dAACgRMxI\nAQAAAIBBBCkAAAAAMIggBQAAAAAGEaQAAAAAwCCCFAAAAAAYRJACAAAAAIMIUgAAAABgEM+RAgBc\nFa7350Zd/vpqfbamGnsCAKgMzEgBAAAAgEEEKQAAAAAwiCAFAAAAAAYRpAAAAADAIBabAADgGnLl\nohwsXAEA1YMZKQAAAAAwiBkpAACqWVXNMlWk3dL2ZVYMAAhSAIBqcr0/NwoAcH0jSAEA4GDVFSKN\nzDJVVrsAcL0iSAEAcINiVhAAyo8gBQAAykToAgBrBCkAAK5iZQWYy8u5pA4AHIcgBQDAdYJZIwBw\nHIIUAMBh+KBvn2t9nJglA3Aj4IG8AAAAAGAQQQoAAAAADOLSPgBApeF5QgCAGwVBCgAAVJli93ut\n2lY9HQGASkaQAgBUmWt90QRUvhP9g61+ZtYSwLWKe6QAAAAAwCCCFAAAAAAYxKV9AAAAVxkjC7ew\nyAtQPQhSAIBScZ8TqlJlhYBrMUxU5MHF1/Pfy6r6XV6L5wiuboaC1OnTp7V06VKZTCZlZWXJ09NT\nHTt2VFhYmNzd3e1u5/z581q+fLkSEhJ05swZeXh4KCgoSIMGDZKXl1eVHhsArjZW/7hX04pmFflA\nB1Sm0s7FinwQLi14VOScr6xAcz0Ho6tVVZ0TuHHYHaQyMjI0efJkZWZmqkOHDvL399ehQ4cUExOj\nvXv3atq0afLw8CiznaysLE2aNEnp6ekKDAxUcHCw0tLSFBcXp8TERE2fPl2NGjWqkmMDwI3EyAdS\n4GpU1nlaWrmRc9xI3er6gF1Vga2094arJUwY6VNV9Z/ZLNhid5CaN2+eMjMzNWzYMIWGhlq2L1y4\nUOvWrVNkZKRGjhxZZjuRkZFKT09Xnz59NHToUMv2mJgYLViwQHPnztXEiROr5NgAAADXuys/9J9w\n0HFKU1WX51UVIyGeUHXjsitIZWRkyGQyycfHRz169LAqGzhwoGJjY7VlyxYNHTpUrq6uJbaTk5Oj\n+Ph4ubi4aMCAAVZlPXv21Lp162QymXTixAnLrFRlHftaU5GbTEvb71q4efV6uza6qi5TudZdjeei\no+7VKO3vrJFn7FTVBwpmq4CSXW9/P6ormFz+3natv5eV9dDpyuqHkdm4irZlrxv5c4xkZ5BKSkqS\nJAUFBalmTesV093c3NSqVSuZTCalpKTozjvvLLGd5ORk5eXlKSgoSG5ublZlNWvWVFBQkGJjY5WU\nlGQJUpV17KtdRS5fqI52r9y3sq5TL7O+gTenivTJSOCpqstHKvJhvCL9v3Lfyz/YV+XlYZU1jlXV\nx8r6u+LIfauiHQA3jqp837ie35Ou/EKsslTXv7kVadfI56VrMYTZFaR+//13SZKfn5/Ncl9fX5lM\nJqWnp5caZuxp5/J6lXnsq9G19iZSWdeiO0pVXltfHd+eVWZgK+9xrsbf85WuhX9oAAC4EVzv/47a\nFaQuXrwoSapdu7bNcvP2CxcuVEo75nqVcezY2FjFxsZKkmbMmKHGjRuX2sfyKHeb63ZXbkduEFbj\nzRgCAADc8KriM35ZapZd5drWvXt3zZgxQzNmzKiS9sePH18l7cI2xtvxGHPHYrwdjzF3LMbb8Rhz\nx2K8Ha+6xtyuIGVrpuhy5u116tSplHYun32qrGMDAAAAQGWxK0iZp8rS09NtlmdkZEgq+T4mo+1c\nPjVXWccGAAAAgMpiV5Bq06aNJMlkMqmwsNCqLDs7WwcPHpSLi4tatGhRajstW7aUs7OzDh48qOzs\nbKuywsJCmUwmq+NV5rGrSvfu3avluDcqxtvxGHPHYrwdjzF3LMbb8Rhzx2K8Ha+6xtyuIOXr66ug\noCCdPHlSGzZssCqLiopSbm6uunbtavUcp7S0NKWlpVnVdXV1Vbdu3ZSbm6tly5ZZla1fv14nT55U\nUFCQZenz8h7bkfjL4liMt+Mx5o7FeDseY+5YjLfjMeaOxXg7XnWNeY2ioqIieypmZGRo8uTJyszM\nVIcOHdSkSROlpKQoKSlJfn5+mj59ujw8PCz1Bw4cKOlS2LlcVlaWJk2apPT0dAUGBqp58+Y6fvy4\ndu/erXr16mnatGmWZdDLe2wAAAAAqEp2BylJOnXqlKKiorR3715lZWXJ09NTnTp1UlhYmNzd3a3q\nlhSkJOn8+fNatmyZEhISdObMGXl4eKht27YaNGiQvLy8KnxsAAAAAKhKhoLUjSI/P188ETtNAAAQ\nTElEQVTffvutUlNTdeTIER0/flwFBQV6/vnn9dBDD5W6b1xcnDZs2KDjx4+rZs2auvXWW/XII4/o\n7rvvtlm/sLBQMTExiouLU3p6upydndWyZUs99thjuv3226vi5V1TPvroI23evLnUOoGBgZoyZYrl\n57i4OH388ccl1n/uuef0t7/9rdL6eL35448/NHr06BLLg4ODNWbMGJtlRs9/XFpIZ+fOnTKZTMrI\nyNDZs2fl7u6uFi1aqFevXgoMDCy2D+d4xZ0+fVpLly6VyWSyfDnXsWNHvpwrp6ysLO3atUt79uzR\n0aNH9eeff8rJyUlNmzbVAw88oJCQENWs+b+7CSryPoNLXnzxRZ08edJmWb169fTZZ58V2/7LL79o\n5cqVSk5OVl5envz8/PTAAw8oNDTU6veD4sp635WkGjVqaOnSpZI4x+21Y8cOHThwQKmpqfrtt9+U\nnZ2t++67Ty+99FKJ+5TnPP7xxx+1du1aHTlyRIWFhbr55pv1t7/9TSEhIRXqv10P5L3R5ObmasGC\nBZIuvRnVr19fp0+fLnO/L774QtHR0fLy8tJDDz2k/Px8bdu2Te+++66effZZ9ezZ06p+UVGR3n//\nfe3YsUONGzdWz549df78eW3btk0mk0njxo1Tx44dq+IlXjM6duwoHx8fm2VbtmzRiRMn1LZtW5vl\nHTp0UEBAQLHtt912W2V28bp1yy232Dz/mjZtarO+0fMflyxdulTbtm1TkyZN1K5dO7m7u+v333/X\n7t27tXv3bv39739Xr169bO7LOV4+V14u7u/vr0OHDikmJkZ79+7VtGnTuFzcoO3bt2vu3Lny9PRU\nmzZt5O3trbNnz2rXrl2aM2eOEhMTFR4erho1aljtZ/R9BtZq165t8/3B1n3jCQkJmjVrlm666SYF\nBwfL3d1dP/74oxYuXKhffvlF4eHhjujyNSsgIEBhYWE2yw4ePKj9+/erXbt2xco4x0u3YsUK/fbb\nb3J1dZWXl1ex9RWuVJ7zeP369fr888/l4eGhrl27ysnJSTt37tTHH3+so0ePaujQoeXuP0HKBhcX\nF02YMEEBAQHy9PRUVFSUli9fXuo+v/zyi6Kjo9WoUSO98847lm80H330UY0fP16LFi1S+/bt1bBh\nQ8s+W7du1Y4dO3T77bdr8uTJcnZ2liQ9/PDDmjJlij799FMFBgbKzc2t6l7sVa5Tp07q1KlTse0X\nLlzQmjVr5OTkVOK3CZ06darwNw03soCAAMslumUpz/mPS9q2bau+ffvq1ltvtdp+4MABTZs2TYsX\nL1aXLl3k6elZbF/O8fKZN2+eMjMzNWzYMIWGhlq2L1y4UOvWrVNkZKRGjhxZjT289jRu3Fivvvqq\n2rdvb/WN8JNPPqkJEyZo586d2rlzpzp37my1n5H3GRRXp04du8bv4sWL+vTTT1WzZk298cYbli9b\nBg0apLfeeks7duzQ1q1bde+991Z1l69ZAQEBNr+4kqSJEydKsr3gAed46Z555hl5eXnJ19dXBw4c\n0Jtvvlli3fKcx3/88YcWLVokd3d3vfPOO5bPIWFhYZowYYKio6PVuXNntWzZslz9Zx7XBicnJ7Vr\n187mB5eSbNy4UZL02GOPWV0W0rBhQ/Xo0UN//fWX4uLibO4zaNAgS4iSpObNmys4OFjnzp3Tjh07\nKvBKrl/x8fHKy8tTp06dVLdu3eruzg2vPOc/LgkJCSkWoiSpdevWatOmjfLz8/XLL79UQ8+uTxkZ\nGTKZTPLx8VGPHj2sygYOHCgXFxdt2bJFOTk51dTDa1NgYKA6dOhQ7LKa+vXr6+GHH5Z06csBVI8d\nO3bo3LlzCg4OtpqxdnZ21hNPPCFJ+vbbb6ure9e0o0ePKiUlRQ0aNFD79u2ruzvXnMDAQPn5+RWb\nrbalPOfxpk2b9Ndff6lnz55WX+a6u7urf//+NvcxgiBVSfbv3y9JNi8zM0/1mutIUl5enn755Re5\nuLjojjvuKLaPuZ3L98H/fPfdd5JKX+4yNTVV69at0+rVqxUfH2/X5Zn4nzNnzmjjxo1auXKlNm7c\nqN9++63EukbPf9inVq1aVv+/Eue4cUlJSZKkoKCgYh/63dzc1KpVK+Xm5iolJaU6unddcnK6dPGL\nrXsXjLzPoLi//vpL8fHxWrlypWJiYrR///5iz9yUSn+PvuOOO+Ti4qLk5GT99ddfVd7n601sbKwk\n6cEHH+Qcr2LlOY/t+Xxi/nehPP6/9u4spoktjAP4HxCLSy+RYqFaXEgAgbjgDioKD0oMauL6Yozx\nQROXuOKTVmKMD5po4vpiXKNRNGJilSWKxAURiYqKrC5YZKm7QHVaWu4Dab29XegMlXuB/++tpzNl\nGL4c+p055zuc2ucFv379wpcvXxAQEOD0KZa1nHt9fb2trbGxERaLBUql0umXJJVK5XAOtausrMT7\n9++hUqmcLsS3unnzpt1rX19fJCcnY+XKlXZPAMm558+f4/nz53ZtsbGxWLduHYKDg21tUuKfOvbx\n40e8fPnS5WALwBiXoq6uDsDvPvbfQkNDUVJSgvr6eowePborL61HMpvNtoJBzr7IeNrPkHPfvn3D\nkSNH7NqUSiXWrl2LmJgYW5u1/x0yZIjDZ/j5+UGpVEKn06GxsRFqtfrPXnQPYjQace/ePVvf6wxj\n3HukxLG7Pn/QoEGQyWT4/PkzBEGATCYTfU1MpLzAYDAAaF/06Yy1vaWlpVPnUDvr6I+rCopKpRKr\nVq3CmDFjoFAoYDAYUF5ejgsXLuDWrVv4+fMnNm7c2JWX3K3IZDIsWrQIkyZNsm2OXVNTg8uXL6O0\ntBS7d+/Gvn37bIuZGcveZzKZcOjQIZhMJixfvtyhihxjXDrGa9c6f/48dDod4uLi7BIpsf0MOZo1\naxaio6OhVqvRr18/NDY2Ijs7G7dv38bevXuxZ88e25oeT+Peehx5pqCgAC0tLRg/frxDUsQY9z4p\ncezJOYIgwGAwMJH6J3dlQZ3pqNQiSefNv4XBYMDDhw/dFpmIiYmxG4mTyWSIj49HREQE0tLS8ODB\nAyxYsMDlotGeoDP3PDAwEMuWLbN7PyYmBjt27IBGo0FVVRXy8vJcVpLrjbwZ4xaLBYcPH0ZFRQUS\nEhIwb948h2MY49Qd3Lx5E1qtFkOHDsWGDRvs3mM/03lLliyxez1s2DCsXr0aAQEB0Gq1uHz5MtLS\n0v6jq+sd3C0zYIz3Dj02kQoJCYG/v7/HxwcFBUn+WR2N5FjbBwwY0Klzuitv/i3u3r0LQRCQkJAg\nushEcHAw4uLicP/+fZSVlfXoL5l/Iv79/PyQnJyMqqoqvHr1ytb596ZYdsVb99tiseDQoUMoLCxE\nfHw8NmzY4NECXKveFONSMV67RnZ2Nk6fPg21Wg2NRuPx3lyu+hny3OzZs6HValFWVmZr8zTuXY3a\nkyOdToeKigooFApRRSYY49JJieP+/fujqakJBoPB6bYWnY39HptI/XOD1j8tICAAQUFB+PLlC75+\n/eqwTqShoQGA/fzMkJAQ+Pr6Qq/Xw2w2O6yTss4DdTWPvzvx5t/COvpjrQIlljX5EgTBa9f0f/Sn\n4t/Z/ZMS/z2NN+53a2urLYmaPn061q9fL2mDzN4S41JZ59a7WrPXG+L1T7tx4wbOnDmDsLAwaDQa\nBAYGijqfMdw5zu6fSqXC69evUVdXh/DwcLvjzWYz9Ho9/Pz8bFPQqGPWZQZJSUmi+2rGuDRS4njI\nkCGoqKhAfX29QyL19etXCIIAhUIhaVofwKp9XmMtevDs2TOH954+fWp3DNBeqjEqKgqCINiNGllZ\nP8ddMYXepqqqCjU1NVCpVIiNjZX0GdXV1QDA/YwkslYy+/c/W7HxT/ZaW1tx4MABFBYWIjExUXIS\nBTDGO2LtO0pKShyqm/38+RPl5eWQyWSIiIj4Ly6v27t27RrOnDmDESNGYNeuXaKTKMB1P0Oeqays\nBGDfB7jro8vKyiAIAiIjI0U9We/NjEYj7t6967bIhDuMcWmkxLEn30+kfqcEmEh5jfUJydWrV9Hc\n3Gxr1+v1yMnJgb+/v8OaHus5ly5dgtFotLVXV1ejoKAAf/31F6ZMmfLnL76bsI7+uCt5DgCvX792\naLNYLMjMzERlZSXkcrnT6lHU7s2bN07L57548QI3btwAAMyYMcPuPSnxT+1MJhP279+P4uJiJCcn\nY+3atR0mUYxx6UJDQzF27Fh8/PgROTk5du9lZGRAEATMmDGDC8AluHLlCi5cuIDw8HBoNBq306+l\n9DP0W21trdO9zvR6PU6ePAnA/v5NnToVcrkcBQUFdv2H0WjExYsXAbRPCSTPFBYWoqWlBePGjXNZ\neY8x7n1S4jgpKQn+/v7Izs6GXq+3tTc3NyMzM9PpOWL4tLW1tUk+uwe7du0aPnz4AKB9r5aamhpE\nRUXZSjmPGjXKoWrc2bNnodVqoVAoMGXKFLS2tuLhw4doamrCqlWrkJKSYnd8W1sbDh48iMLCQgwd\nOhQTJkxAU1MTCgoKYDKZsHXrVkyaNKlrfuH/OYPBgDVr1sBiseD48eNu/0EvXboUYWFhGD58OIKC\ngmAwGFBRUQGdTgeZTIZt27Zh7NixXXj13Ut6ejrq6+sRFRVlW8vz/v17214My5Ytw6JFixzOExv/\n1O7YsWPIz8+HXC532CDWKjY21m7EjDHeOQ0NDdi5cye+f/+OiRMnQq1Wo6qqCqWlpVCpVNizZ4/T\nufTkWn5+Po4dOwZfX1+kpKQ4XW+gVCptAypS+xlql5GRAa1Wi+joaAwePBgBAQFobGzEkydPYDKZ\nEBcXh7S0NNseXgBQVFSEAwcOwN/fH9OmTcPAgQNRXFyMuro6TJ06FZs3bxa1JrM302g0KC8vx/bt\n2zFx4kSnxzDGPVNUVITHjx8DaC/nX1JSgpCQEIwaNQoAIJfLsWLFCrvjxcZxVlYWTp06Bblcjvj4\nePTp0wePHj3C58+fkZqaavf5YjGRciE9Pd3tLuwzZ87EunXrHNrz8/ORk5OD2tpa+Pj4YOTIkZg/\nfz4mTJjg9HPMZjOysrJw584dNDQ0oG/fvoiMjMTChQsRFRXltd+nu8vNzcWJEyeQkJCATZs2uT32\n3LlzqK6uRkNDA5qbm+Hj44Pg4GCMHj0aqampfJTegby8PBQVFUGn0+HHjx8wm80IDAxEZGQkUlJS\nXO5pBIiPf+q4rwGAxYsXY+nSpbbXjPHO+/TpEzIyMvDs2TM0NTVh0KBBmDx5MhYvXuxxYQT6LSMj\nA1euXHF7TExMDNLT0wF0rp8h4NWrV8jNzcW7d+/w7ds3CIKA/v37Y8SIEUhMTERiYqLTpKi8vNz2\n5NpoNCI0NBRJSUmYO3eu5OnEvU1tbS22bNkChUKBo0ePurxvjHHPdNR3DB48GEePHrVrkxLHxcXF\nuH79Ot6+fYu2tjao1WrMmTOn07NlmEgRERERERGJxOEHIiIiIiIikZhIERERERERicREioiIiIiI\nSCQmUkRERERERCIxkSIiIiIiIhKJiRQREREREZFITKSIiIiIiIhEYiJFREREREQkEhMpIiIiIiIi\nkZhIERERERERifQ3bZnvwzn0R84AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fcc0c3a7cd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# generate exclusions *except* early/late suspicion\n",
    "COL_REM = ['exclusion_nonadult', 'exclusion_secondarystay',\n",
    "           'exclusion_csurg','exclusion_carevue',\n",
    "           #'exclusion_early_suspicion', 'exclusion_late_suspicion',\n",
    "           'exclusion_bad_data']\n",
    "idxRem = np.zeros(co.shape[0])\n",
    "for c in COL_REM:\n",
    "    N_REM = np.sum( (idxRem == 0) & (co[c].values==1) )\n",
    "    idxRem[co[c].values==1] = 1\n",
    "    \n",
    "idxRem = idxRem.astype(bool)\n",
    "\n",
    "# distribution of time of infection\n",
    "N_HR = 96\n",
    "xi = np.linspace(-N_HR, N_HR, N_HR*2+1)\n",
    "col = gg_color_hue(5)\n",
    "\n",
    "pretty_dict = {'suspected_infection_time_poe_days': 'Blood culture + ABX'}\n",
    "c = 'suspected_infection_time_poe_days'\n",
    "\n",
    "idxKeep = ~co[c].isnull()\n",
    "tmp = co.loc[~idxRem & idxKeep, c].values * 24.0\n",
    "N_firstday = sum( (tmp>-24) & (tmp<24) )\n",
    "\n",
    "lbl_str = pretty_dict[c]\n",
    "lbl_str += '\\n' + '{:2.2f}% of grp >  24hr.'.format(sum(tmp>24)*100.0 / np.sum(~idxRem & idxKeep))\n",
    "lbl_str += '\\n' + '{:2.2f}% of grp  +-24hr.'.format(sum((tmp>=-24) & (tmp<=24))*100.0 / np.sum(~idxRem & idxKeep))\n",
    "lbl_str += '\\n' + '{:2.2f}% of grp < -24hr.'.format(sum(tmp<-24)*100.0 / np.sum(~idxRem & idxKeep))\n",
    "\n",
    "plt.figure(figsize=[14,6])\n",
    "plt.hist( tmp, bins=xi, label=lbl_str, normed=True )\n",
    "plt.title('{} ({:2.2f}%) in [-24,24]. '.format(\n",
    "        N_firstday, N_firstday*100.0 / np.sum(~idxRem)))\n",
    "plt.legend(loc='upper left')\n",
    "    \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that most patients are suspected of infection either before, or at the time of their ICU admission. This motivates the decision to evaluate the performance of the scores at ICU admission."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load final dataset from Postgres"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load in final dataset - note we apply the exclusion criteria with excluded=0\n",
    "query = query_schema + \"select * from sepsis3 where excluded = 0\"\n",
    "\n",
    "df = pd.read_sql_query(query,con)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have: ICU intime/outtime, suspected infection time, whether the microbiology culture was positive, some demographics, comorbidities, outcomes, and the severity scores. \n",
    "\n",
    "The severity scores are extracted at a [0, 24] hour window centered around ICU admission - except labs have an extended [-6, 24] hour window (i.e. 'sofa' is extracted in this way)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Add in useful variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# add the composite outcome\n",
    "df['composite_outcome'] = ( (df['hospital_expire_flag']==1) | (df['icu_los']>=3) ).astype(int)\n",
    "\n",
    "labels = OrderedDict([['suspicion_poe', 'BC + ABX (Prescribed)']])\n",
    "\n",
    "# add some other useful variables\n",
    "df['blood culture'] = (~df['blood_culture_time'].isnull())\n",
    "df['suspicion_poe'] = (~df['suspected_infection_time_poe_days'].isnull())\n",
    "\n",
    "df['abx_poe'] = (~df['antibiotic_time_poe'].isnull())\n",
    "\n",
    "df['sepsis-3'] = ((df['suspicion_poe']==1) & (df['sofa']>=2)).astype(int)\n",
    "df['sofa>=2'] = (df['sofa']>=2).astype(int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Baseline model + scores\n",
    "\n",
    "The original paper evaluates a *baseline model* with the addition of the various severity scores. \n",
    "\n",
    "> To measure predictive validity, a baseline risk model was created for in-hospital mortality based on preinfection criteria using multivariable logistic regression. The baseline model included age (as a fractional polynomial), sex, race/ethnicity (black, white, or other), and the weighted Charlson comorbidity score (as fractional polynomial) as a measure of chronic comorbidities.\n",
    "\n",
    "This baseline model includes:\n",
    "\n",
    "* age (fractional polynomial)\n",
    "* sex\n",
    "* ethnicity\n",
    "* Charlson comorbidities (fractional polynomial)\n",
    "\n",
    "We can reproduce this model (done in the appendix), with the following caveats:\n",
    "\n",
    "1. We will build and evaluate the model on the same dataset, so our estimates are \"apparent\"\n",
    "2. We will use Elixhauser comorbidities, not Charlson comorbidities\n",
    "3. We may not have identical fractional polynomial terms (as we are rebuilding the model on our dataset)\n",
    "\n",
    "The following code block extracts the covariates for the baseline model. R code in the repository can use this file to build a fractional polynomial model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_header = ['age','elixhauser_hospital','hospital_expire_flag','sepsis_angus',\n",
    "            'is_male','race_black','race_other',\n",
    "            'qsofa','sofa','sirs','lods']\n",
    "\n",
    "X = df[X_header].values\n",
    "\n",
    "# we'll write out the design matrix for the MFP model here - this is used by the R code\n",
    "np.savetxt(os.path.join(data_path, 'sepsis3-design-matrix.csv'), X, fmt='%4.4f',\n",
    "           delimiter=',', header=','.join(X_header), comments='')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Save the data to file\n",
    "\n",
    "The dataframes will be loaded directly from a file, rather than the database.\n",
    "\n",
    "The `X` data which was written to 'sepsis3-design-matrix.csv' will be used by the `print_auc_table_baseline` function to evaluate the AUROC of the scores when incorporated with the baseline model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv(os.path.join(data_path, 'sepsis3-df.csv'),sep=',',index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dataset with no exclusions\n",
    "\n",
    "It may be useful for others to analyze the dataset without exclusions. Here we generate an identical copy of the data, except for all `icustay_id` in MIMIC-III."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load in final dataset - note we add in the individual exclusion criteria\n",
    "query = query_schema + \"\"\"\n",
    "select ie.subject_id\n",
    ", s.*\n",
    ", co.exclusion_secondarystay\n",
    ", co.exclusion_nonadult\n",
    ", co.exclusion_csurg\n",
    ", co.exclusion_carevue\n",
    ", co.exclusion_early_suspicion\n",
    ", co.exclusion_late_suspicion\n",
    ", co.exclusion_bad_data\n",
    "from sepsis3 s\n",
    "-- add in subject_id\n",
    "inner join icustays ie\n",
    "  on s.icustay_id = ie.icustay_id\n",
    "inner join sepsis3_cohort co\n",
    "  on s.icustay_id = co.icustay_id\n",
    "order by s.icustay_id\n",
    "\"\"\"\n",
    "\n",
    "df = pd.read_sql_query(query,con)\n",
    "\n",
    "# add the composite outcome\n",
    "df['composite_outcome'] = ( (df['hospital_expire_flag']==1) | (df['icu_los']>=3) ).astype(int)\n",
    "\n",
    "labels = OrderedDict([['suspicion_poe', 'BC + ABX (Prescribed)']])\n",
    "\n",
    "# add some other useful variables\n",
    "df['blood culture'] = (~df['blood_culture_time'].isnull())\n",
    "df['suspicion_poe'] = (~df['suspected_infection_time_poe_days'].isnull())\n",
    "\n",
    "df['abx_poe'] = (~df['antibiotic_time_poe'].isnull())\n",
    "\n",
    "df['sepsis-3'] = ((df['suspicion_poe']==1) & (df['sofa']>=2)).astype(int)\n",
    "df['sofa>=2'] = (df['sofa']>=2).astype(int)\n",
    "\n",
    "df.to_csv(os.path.join(data_path, 'sepsis3-df-no-exclusions.csv'),sep=',',index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "con.close()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "sepsis3-py3",
   "language": "python",
   "name": "sepsis3-py3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}