Pathway patterns throughout the day

Aims

Show key pathway statistics broken down by time of day (3 hour epochs).

Import libraries and data

Data has been resticted to stroke teams with at least 300 admissions, with at least 10 patients receiving thrombolysis, over three years.

# import libraries
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

# Display entire dataframes
pd.set_option("display.max_rows", 999, "display.max_columns", 150)

# import data
raw_data = pd.read_csv(
    './../data/2019-11-04-HQIP303-Exeter_MA.csv', low_memory=False)

headings = list(raw_data)
print (headings)

['StrokeTeam', 'PatientUID', 'Pathway', 'S1AgeOnArrival', 'MoreEqual80y', 'S1Gender', 'S1Ethnicity', 'S1OnsetInHospital', 'S1OnsetToArrival_min', 'S1OnsetDateType', 'S1OnsetTimeType', 'S1ArriveByAmbulance', 'S1AdmissionHour', 'S1AdmissionDay', 'S1AdmissionQuarter', 'S1AdmissionYear', 'CongestiveHeartFailure', 'Hypertension', 'AtrialFibrillation', 'Diabetes', 'StrokeTIA', 'AFAntiplatelet', 'AFAnticoagulent', 'AFAnticoagulentVitK', 'AFAnticoagulentDOAC', 'AFAnticoagulentHeparin', 'S2INR', 'S2INRHigh', 'S2INRNK', 'S2NewAFDiagnosis', 'S2RankinBeforeStroke', 'Loc', 'LocQuestions', 'LocCommands', 'BestGaze', 'Visual', 'FacialPalsy', 'MotorArmLeft', 'MotorArmRight', 'MotorLegLeft', 'MotorLegRight', 'LimbAtaxia', 'Sensory', 'BestLanguage', 'Dysarthria', 'ExtinctionInattention', 'S2NihssArrival', 'S2BrainImagingTime_min', 'S2StrokeType', 'S2Thrombolysis', 'Haemorrhagic', 'TimeWindow', 'Comorbidity', 'Medication', 'Refusal', 'Age', 'Improving', 'TooMildSevere', 'TimeUnknownWakeUp', 'OtherMedical', 'S2ThrombolysisTime_min', 'S2TIAInLastMonth']

Restrict original data to hospitals with at least 300 admissions + 10 thrombolysis patients

keep = []

groups = raw_data.groupby('StrokeTeam') # creates a new object of groups of data

for index, group_df in groups: # each group has an index and a dataframe of data
    
    # Skip if total admissions less than 300 or total thrombolysis < 10
    admissions = group_df.shape[0]
    thrombolysis_received = group_df['S2Thrombolysis'] == 'Yes' 
    if (admissions < 300) or (thrombolysis_received.sum() < 10):
        continue
    
    else: 
        keep.append(group_df)

# Concatenate output
data = pd.DataFrame()
data = pd.concat(keep)

Remove in-hospital onset

mask = data['S1OnsetInHospital'] == 'No'
data = data[mask]

Group by time of day

# Work on copy of data
data_time = data.copy()

# Set up results DataFrame
time_summary = pd.DataFrame()

# Count arrivals
time_summary['Arrivals'] = \
    data_time.groupby('S1AdmissionHour').count()['StrokeTeam']

# Get thrombolysis rate
thrombolysed = data_time['S2Thrombolysis'] == 'Yes'
data_time['thrombolysed'] = thrombolysed
time_summary['thrombolyse_all'] = \
    data_time.groupby('S1AdmissionHour').mean()['thrombolysed']

# Get proportion of strokes with known onset
onset_known = (data_time['S1OnsetTimeType'] == 'Best estimate') | \
    (data_time['S1OnsetTimeType'] == 'Precise')
data_time['onset_known'] = onset_known
time_summary['onset_known'] = \
    data_time.groupby('S1AdmissionHour').mean()['onset_known']

# Get proportion over 80
data_time['age_80_plus'] = data_time['MoreEqual80y'] == 'Yes'
time_summary['age_80_plus'] = \
    data_time.groupby('S1AdmissionHour').mean()['age_80_plus']

# Get Rankin score
time_summary['rankin_all'] = \
    data_time.groupby('S1AdmissionHour').mean()['S2RankinBeforeStroke']

# Get NIHSS
time_summary['nihss_all'] = \
    data_time.groupby('S1AdmissionHour').mean()['S2NihssArrival']

# Get onset to arrival <4hrs and then restrict data
data_time['4hr_arrival'] = data_time['S1OnsetToArrival_min'] <= 240
time_summary['4hr_arrival'] = \
    data_time.groupby('S1AdmissionHour').mean()['4hr_arrival']
mask = data_time['4hr_arrival']
data_time = data_time[mask]

# Get Rankin score of arrivals within 4hrs onset
time_summary['rankin_4hr'] = \
    data_time.groupby('S1AdmissionHour').mean()['S2RankinBeforeStroke']

# Get NIHSS of arrivals within 4hrs onset
time_summary['nihss_4hr'] = \
    data_time.groupby('S1AdmissionHour').mean()['S2NihssArrival']

# Get onset to arrival (of those arriving within 4 hours)
time_summary['onset_arrival'] = \
    data_time.groupby('S1AdmissionHour').mean()['S1OnsetToArrival_min']

# Get scan in four hours (and remove rest)
data_time['4hr_scan'] = data_time['S2BrainImagingTime_min'] <= 240
time_summary['scan_4hrs'] = \
    data_time.groupby('S1AdmissionHour').mean()['4hr_scan']
mask = data_time['4hr_scan']
data_time = data_time[mask]

# Get arrival to scan (of those arriving within 4 hours and scanned in 4hrs)
time_summary['arrival_scan'] = \
    data_time.groupby('S1AdmissionHour').mean()['S2BrainImagingTime_min']

# Filter down to acanned within 4 hrs onset
onset_to_scan = (
    data_time['S1OnsetToArrival_min'] + data_time['S2BrainImagingTime_min'])
data_time['onset_to_scan'] = onset_to_scan
mask = data_time['onset_to_scan'] <= 240
data_time = data_time[mask]

# Get thrombolysis given and remove rest
time_summary['thrombolyse_4hr'] = \
    data_time.groupby('S1AdmissionHour').mean()['thrombolysed']
mask = data_time['thrombolysed']

# Get scan to needle
scan_to_needle = (
    data_time['S2ThrombolysisTime_min'] - data_time['S2BrainImagingTime_min'])
data_time['scan_needle'] = scan_to_needle
time_summary['scan_to_needle'] = \
    data_time.groupby('S1AdmissionHour').mean()['scan_needle']

# Get arrival to needle
time_summary['arrival_to_needle'] = \
    data_time.groupby('S1AdmissionHour').mean()['S2ThrombolysisTime_min']

# Get onset to needle 
onset_to_needle = (
    data_time['S1OnsetToArrival_min'] + data_time['S2ThrombolysisTime_min'])
data_time['onset_to_needle'] = onset_to_needle
time_summary['onset_to_needle'] = \
    data_time.groupby('S1AdmissionHour').mean()['onset_to_needle']

# Add short time start (string) column for plots
hour = ['0', '3', '6', '9', '12', '15', '19', '21']
time_summary['hour'] = hour

Show summary table by time of day

time_summary.T

S1AdmissionHour	00:00 to 02:59	03:00 to 05:59	06:00 to 08:59	09:00 to 11:59	12:00 to 14:59	15:00 to 17:59	18:00 to 20:59	21:00 to 23:59
Arrivals	9890	7367	15871	52049	48687	39838	32200	20918
thrombolyse_all	0.131446	0.0651554	0.109004	0.119464	0.118882	0.115945	0.119658	0.135338
onset_known	0.75996	0.679381	0.636633	0.63006	0.646004	0.658241	0.708571	0.75251
age_80_plus	0.378564	0.379938	0.387562	0.436589	0.423768	0.414027	0.411677	0.412037
rankin_all	1.02053	1.0566	1.05003	1.07894	1.03551	1.02995	1.03165	1.06062
nihss_all	7.94191	8.35454	8.80986	7.71318	7.0694	6.80621	7.38032	7.62235
4hr_arrival	0.464307	0.353197	0.388066	0.378778	0.36938	0.365304	0.423137	0.46362
rankin_4hr	0.93162	1.04343	1.01153	1.13178	1.06611	1.02185	1.07134	1.03722
nihss_4hr	9.47393	9.53222	10.0548	9.30759	8.91874	8.91479	9.32159	9.1475
onset_arrival	122.127	115.238	102.11	108.779	113.879	111.436	110.78	112.673
scan_4hrs	0.922474	0.895465	0.958597	0.95861	0.95863	0.954786	0.939523	0.926892
arrival_scan	40.8057	47.2502	44.7102	41.357	42.0989	40.9557	42.8849	41.8893
thrombolyse_4hr	0.334958	0.223671	0.310397	0.360533	0.373564	0.366626	0.333631	0.349285
scan_to_needle	44.0687	44.2937	39.3118	31.7291	32.1522	33.7953	40.7675	44.3519
arrival_to_needle	68.2312	68.0112	63.5706	53.7009	53.7034	55.1399	63.8725	69.2026
onset_to_needle	177.516	175.269	150.697	146.731	150.852	149.811	158.123	165.162
hour	0	3	6	9	12	15	19	21

Show summary charts of key metrics by time of day

# Set up figure
fig = plt.figure(figsize=(12,15))

# Subplot 1: Arrivals
ax1 = fig.add_subplot(4,4,1)
x = time_summary['hour']
y = time_summary['Arrivals'] / time_summary['Arrivals'].sum() * 8
ax1.plot(x,y)
# Add line at 1
y1 = np.repeat(1,8)
ax1.plot(x,y1, color='0.5', linestyle=':')
ax1.set_ylim(ymin=0) # must be after plot method
ax1.set_xlabel('Time of day')
ax1.set_ylabel('Normlalised arrivals')
ax1.set_title('Arrivals\n(normalised to average)')

# Subplot 2: Thrombolysis
ax2 = fig.add_subplot(4,4,2)
x = time_summary['hour']
y = time_summary['thrombolyse_all'] * 100
ax2.plot(x,y)
ax2.set_xlabel('Time of day')
ax2.set_ylabel('Thrombolysis (%)')
ax2.set_title('Thrombolysis use\n(all arrivals)')

# Subplot 3: Known onset
ax3 = fig.add_subplot(4,4,3)
x = time_summary['hour']
y = time_summary['onset_known'] * 100
ax3.plot(x,y)
ax3.set_xlabel('Time of day')
ax3.set_ylabel('Onset known (%)')
ax3.set_title('Proportion with known onset')

# Subplot 4: age_80_plus 
ax4 = fig.add_subplot(4,4,4)
x = time_summary['hour']
y = time_summary['age_80_plus'] * 100
ax4.plot(x,y)
ax4.set_xlabel('Time of day')
ax4.set_ylabel('Aged 80+')
ax4.set_title('Proportion aged 80+')

# Subplot 5: Rankin (all arrivals)
ax5 = fig.add_subplot(4,4,5)
x = time_summary['hour']
y = time_summary['rankin_all'] 
ax5.plot(x,y)
ax5.set_xlabel('Time of day')
ax5.set_ylabel('Rankin (mean)')
ax5.set_title('Mean pre-stroke\nmodified Rankin\n(all arrivals)')

# Subplot 6: NIHSS (all arrivals)
ax6 = fig.add_subplot(4,4,6)
x = time_summary['hour']
y = time_summary['nihss_all'] 
ax6.plot(x,y)
ax6.set_xlabel('Time of day')
ax6.set_ylabel('NIHSS (mean)')
ax6.set_title('Mean NIH Stroke Scale\n(all arrivals)')

# Subplot 7: 4hr_arrival
ax7 = fig.add_subplot(4,4,7)
x = time_summary['hour']
y = time_summary['4hr_arrival'] * 100
ax7.plot(x,y)
ax7.set_xlabel('Time of day')
ax7.set_ylabel('Arrive within 4hrs of onset (%)')
ax7.set_title('Proportion arriving within\n4hrs of known onset')

# Subplot 8: Rankin (4hr arrivals)
ax8 = fig.add_subplot(4,4,8)
x = time_summary['hour']
y = time_summary['rankin_4hr'] 
ax8.plot(x,y)
ax8.set_xlabel('Time of day')
ax8.set_ylabel('Rankin (mean)')
ax8.set_title('Mean pre-stroke\nmodified Rankin\n(arrivals 4hrs from onset)')

# Subplot 9: NIHSS (4hr arrivals)
ax9 = fig.add_subplot(4,4,9)
x = time_summary['hour']
y = time_summary['nihss_4hr'] 
ax9.plot(x,y)
ax9.set_xlabel('Time of day')
ax9.set_ylabel('NIHSS (mean)')
ax9.set_title('Mean NIH Stroke Scale\n(arrivals 4hrs from onset)')

# Subplot 10: onset_arrival (4hr arrivals)
ax10 = fig.add_subplot(4,4,10)
x = time_summary['hour']
y = time_summary['onset_arrival'] 
ax10.plot(x,y)
ax10.set_xlabel('Time of day')
ax10.set_ylabel('Onset to arrival (minutes, mean)')
ax10.set_title('Mean onset to arrival\n(arrivals 4hrs from onset)')

# Subplot 11: scan_4hrs (4hr arrivals)
ax11 = fig.add_subplot(4,4,11)
x = time_summary['hour']
y = time_summary['scan_4hrs'] * 100
ax11.plot(x,y)
ax11.set_xlabel('Time of day')
ax11.set_ylabel('Proportion scanned within 4hrs (%)')
ax11.set_title('Proportion scanned within\n4hrs of arrival\n(arrivals 4hrs from onset)')

# Subplot 12: arrival_scan (4hr scan)
ax12 = fig.add_subplot(4,4,12)
x = time_summary['hour']
y = time_summary['arrival_scan'] 
ax12.plot(x,y)
ax12.set_xlabel('Time of day')
ax12.set_ylabel('Arrival to scan (minutes, mean)')
ax12.set_title('Mean arrival to scan\n(scanned 4hrs from onset)')

# Subplot 13: thrombolysis (4hr scan)
ax13 = fig.add_subplot(4,4,13)
x = time_summary['hour']
y = time_summary['thrombolyse_4hr']  * 100
ax13.plot(x,y)
ax13.set_xlabel('Time of day')
ax13.set_ylabel('Thrombolsyis(%)')
ax13.set_title('Thrombolysis use\n(scanned 4hrs from onset)')

# Subplot 14: scan_to_needle
ax14 = fig.add_subplot(4,4,14)
x = time_summary['hour']
y = time_summary['scan_to_needle'] 
ax14.plot(x,y)
ax14.set_xlabel('Time of day')
ax14.set_ylabel('Scan to needle (minutes, mean)')
ax14.set_title('Mean scan to needle')

# Subplot 15: arrival_to_needle
ax15 = fig.add_subplot(4,4,15)
x = time_summary['hour']
y = time_summary['arrival_to_needle'] 
ax15.plot(x,y)
ax15.set_xlabel('Time of day')
ax15.set_ylabel('Arrival to needle (minutes, mean)')
ax15.set_title('Mean arrival to needle')

# Subplot 16: onset_to_needle
ax16 = fig.add_subplot(4,4,16)
x = time_summary['hour']
y = time_summary['onset_to_needle'] 
ax16.plot(x,y)
ax16.set_xlabel('Time of day')
ax16.set_ylabel('Onset to needle (minutes, mean)')
ax16.set_title('Mean onset to needle')

# Save and show
plt.tight_layout(pad=2)
plt.savefig('output/stats_by_time_of_day.jpg', dpi=300)
plt.show();

Observations

• Arrivals peak between 9am and 3pm, and then drop steadily to a low between 3am and 6pm.

• Thrombolysis use is significantly lower in the 3am to 6am epoch, but only 3% of admissions occur in this period.

• Arrival-to-scan time is higher 3am to 9am, but reasonably stable for the rest oif the day.

• Scan-to-needle is approximately 10 minutes slower at night (9pm to 9am) than during the day (9am to 9pm).

• The lowest proportion of arrivals with known stroke onset occurs during the day (6am to 6pm).