Association between patient features on use of thrombolysis

Aims

Examine thrombolysis use (in all arrivals, and the subset of those who arrive within 4 hours of known onset) according to:

• Disability (Rankin) before stroke

• Stroke severity (NIHSS)

• Gender

• Ethnicity

• Age group

• Onset known

• Arrival by ambulance

Examine thrombolysis use in the absence and presence of co-morbidities.

Note: The association of particular features with use of thrombolysis does not imply that these relationships are causal.

Examine reasons given for not giving thrombolysis.

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

# 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)

# Convert thromboilysis to boolean
data['thrombolysis_given'] = data['S2Thrombolysis'] == 'Yes'

print (data.shape)

(239505, 63)

Get out-of-hopsital stroke onset arrivals within 4 hours of known stroke onset

# Get out of hospital arrivals
mask = data['S1OnsetInHospital'] == 'No'
data = data[mask]

# Get arrivals within 4 hours of known stroke onset
mask = data['S1OnsetToArrival_min'] <= 240
data_4hr = data[mask]

Breakdown of thrombolysis use by feature

Rankin before stroke

# All admissions
rankin_before_stroke_all = \
    data.groupby('S2RankinBeforeStroke').mean()['thrombolysis_given']
rankin_before_stroke_all.rename('All', inplace=True)

# Admissions within 4 hours of arrival
rankin_before_stroke_4hr = \
    data_4hr.groupby('S2RankinBeforeStroke').mean()['thrombolysis_given']
rankin_before_stroke_4hr.rename('4hr', inplace=True)

# Combine
rankin_before_stroke = \
    pd.concat([rankin_before_stroke_all, rankin_before_stroke_4hr], axis=1)
rankin_before_stroke = rankin_before_stroke.round(3)
rankin_before_stroke *= 100

rankin_before_stroke

	All	4hr
S2RankinBeforeStroke
0	14.2	34.9
1	10.8	28.6
2	9.0	24.0
3	8.7	21.5
4	6.5	15.0
5	4.3	9.8

# Set up figure
fig = plt.figure(figsize=(5,5))
ax = fig.add_subplot()
x = rankin_before_stroke.index
y1 = rankin_before_stroke['All']
y2 = rankin_before_stroke['4hr'] 
ax.plot(x, y1, label='All arrivals')
ax.plot(x, y2, label='4 Hour arrivals')
ax.set_xlabel('Disability (Rankin) before stroke')
ax.set_ylabel('Thrombolysis use (%)')
ax.legend()
plt.tight_layout()
plt.savefig('output/thrombolysis_use_by_rankin_before_stroke.jpg', dpi=300)
plt.show()

Stroke severity

# All admissions
NIHSS_all = data.groupby('S2NihssArrival').mean()['thrombolysis_given']
NIHSS_all.rename('All', inplace=True)

# Admissions within 4 hours of arrival
NIHSS_4hr = data_4hr.groupby('S2NihssArrival').mean()['thrombolysis_given']
NIHSS_4hr.rename('4hr', inplace=True)

# Combine
NIHSS = pd.concat([NIHSS_all, NIHSS_4hr], axis=1) * 100
NIHSS = NIHSS.round(1)

NIHSS

	All	4hr
S2NihssArrival
0.0	0.6	2.1
1.0	1.1	3.8
2.0	2.5	8.3
3.0	4.9	15.0
4.0	9.7	28.7
5.0	14.0	36.8
6.0	16.4	39.9
7.0	18.8	42.4
8.0	21.0	44.7
9.0	21.6	44.2
10.0	21.9	43.4
11.0	24.0	46.1
12.0	23.9	43.8
13.0	24.7	43.9
14.0	25.9	45.9
15.0	25.9	45.0
16.0	26.4	44.4
17.0	26.2	44.3
18.0	26.6	45.1
19.0	26.7	46.2
20.0	23.7	42.3
21.0	24.3	42.6
22.0	24.1	41.4
23.0	25.0	44.2
24.0	24.7	42.7
25.0	22.2	39.3
26.0	17.5	31.0
27.0	17.6	31.8
28.0	12.5	24.1
29.0	12.6	24.0
30.0	9.8	19.4
31.0	8.9	18.1
32.0	8.3	16.5
33.0	6.0	12.7
34.0	4.9	10.0
35.0	4.2	7.7
36.0	2.4	5.0
37.0	3.6	6.8
38.0	5.7	9.9
39.0	6.5	11.4
40.0	0.0	0.0
41.0	0.0	0.0
42.0	1.6	2.7

# Set up figure
fig = plt.figure(figsize=(5,5))
ax = fig.add_subplot()
x = NIHSS.index
y1 = NIHSS['All']
y2 = NIHSS['4hr'] 
ax.plot(x, y1, label='All arrivals')
ax.plot(x, y2, label='4 Hour arrivals')
ax.set_xlabel('Stroke severity (NIHSS) on arrival')
ax.set_ylabel('Thrombolysis use (%)')
ax.legend()

plt.tight_layout()
plt.savefig('output/thrombolysis_use_by_arrival_nihss.jpg', dpi=300)
plt.show()

Show thrombolysis use by NIHSS cutoff of 0-10 or 11+

# All arrivals

mask = (data['S2NihssArrival'] >= 0) & (data['S2NihssArrival'] <= 10)
thrombolisis_use = data[mask]['thrombolysis_given'].mean()
print (
    f'Thrombolysis use in NIHSS 0-10, all arrivals: {thrombolisis_use:0.3f}')

mask = data['S2NihssArrival'] >= 11
thrombolisis_use = data[mask]['thrombolysis_given'].mean()
print (
    f'Thrombolysis use in NIHSS 11+, all arrivals: {thrombolisis_use:0.3f}')

Thrombolysis use in NIHSS 0-10, all arrivals: 0.086
Thrombolysis use in NIHSS 11+, all arrivals: 0.231

# 4 Hour arrivals
mask = (data_4hr['S2NihssArrival'] >= 0) & (data_4hr['S2NihssArrival'] <= 10)
thrombolisis_use = data_4hr[mask]['thrombolysis_given'].mean()
print (
    f'Thrombolysis use in NIHSS 0-10, 4 hour arrivals: {thrombolisis_use:0.3f}')

mask = data_4hr['S2NihssArrival'] >= 11
thrombolisis_use = data_4hr[mask]['thrombolysis_given'].mean()
print (
    f'Thrombolysis use in NIHSS 11+, 4 hour arrivals: {thrombolisis_use:0.3f}')

Thrombolysis use in NIHSS 0-10, 4 hour arrivals: 0.245
Thrombolysis use in NIHSS 11+, 4 hour arrivals: 0.412

Gender

# All admissions
gender_all = data.groupby('S1Gender').mean()['thrombolysis_given']
gender_all.rename('All', inplace=True)

# Admissions within 4 hours of arrival
gender_4hr = data_4hr.groupby('S1Gender').mean()['thrombolysis_given']
gender_4hr.rename('4hr', inplace=True)

# Combine
gender = pd.concat([gender_all, gender_4hr], axis=1) * 100
gender = gender.round(1)

gender

	All	4hr
S1Gender
Female	11.3	28.2
Male	12.3	30.8

Ethnicity

# All admissions
ethnicity_all = data.groupby('S1Ethnicity').mean()['thrombolysis_given']
ethnicity_all.rename('All', inplace=True)

# Admissions within 4 hours of arrival
ethnicity_4hr = data_4hr.groupby('S1Ethnicity').mean()['thrombolysis_given']
ethnicity_4hr.rename('4hr', inplace=True)

# Combine
ethnicity = pd.concat([ethnicity_all, ethnicity_4hr], axis=1) * 100
ethnicity = ethnicity.round(1)

ethnicity

	All	4hr
S1Ethnicity
Asian	12.1	30.0
Black	10.2	29.2
Mixed	11.7	32.3
Other	14.4	34.3
White	11.6	29.1

Age group

# All admissions
age_group_all = data.groupby('S1AgeOnArrival').mean()['thrombolysis_given']
age_group_all.rename('All', inplace=True)

# Admissions within 4 hours of arrival
age_group_4hr = data_4hr.groupby('S1AgeOnArrival').mean()['thrombolysis_given']
age_group_4hr.rename('4hr', inplace=True)

# Combine
age_group = pd.concat([age_group_all, age_group_4hr], axis=1) * 100
age_group = age_group.round(3)

# Re-order
order = ['[15,20)', '[20,25)', '[25,30)', '[30,35)', '[35,40)',
       '[40,45)', '[45,50)', '[50,55)', '[55,60)', '[60,65)', '[65,70)',
       '[70,75)', '[75,80)', '[80,85)', '[85,90)', '[90,95)', '[95,100)',
        '[100,120]']
age_group = age_group.loc[order]
age_group.set_index(np.arange(15, 101, 5), inplace=True)

age_group

	All	4hr
15	17.470	42.029
20	13.402	33.621
25	15.371	37.229
30	14.624	36.164
35	15.697	36.794
40	15.274	37.671
45	14.863	36.277
50	14.189	36.692
55	13.091	34.348
60	12.642	33.238
65	12.843	32.862
70	12.999	32.381
75	12.314	30.396
80	11.298	28.106
85	9.958	24.600
90	8.510	20.752
95	8.630	20.101
100	7.929	17.883

# Set up figure
fig = plt.figure(figsize=(5,5))
ax = fig.add_subplot()
x = age_group.index
y1 = age_group['All']
y2 = age_group['4hr']
ax.plot(x, y1, label='All arrivals')
ax.plot(x, y2, label='4 Hour arrivals')
ax.set_xlabel('Age band (lower bound)')
ax.set_ylabel('Thrombolysis use (%)')
ax.legend()
plt.tight_layout()
plt.savefig('output/thrombolysis_use_by_age.jpg', dpi=300)
plt.show()

Onset known

Get fractions for each onset type

counts = data.groupby('S1OnsetTimeType').count()['StrokeTeam']
fractions = counts/counts.sum() * 100
fractions = fractions.round(1)
fractions

S1OnsetTimeType
Best estimate    33.7
Not known        33.1
Precise          33.1
Name: StrokeTeam, dtype: float64

# All admissions
onset_known_all = data.groupby('S1OnsetTimeType').mean()['thrombolysis_given']
onset_known_all.rename('All', inplace=True)

# Admissions within 4 hours of arrival
onset_known_4hr = data_4hr.groupby('S1OnsetTimeType').mean()['thrombolysis_given']
onset_known_4hr.rename('4hr', inplace=True)

# Combine
onset_known = pd.concat([onset_known_all, onset_known_4hr], axis=1) * 100
onset_known = onset_known.round(1)

# Re-order
order = ['Not known', 'Best estimate', 'Precise']
onset_known = onset_known.loc[order]

onset_known

	All	4hr
S1OnsetTimeType
Not known	0.4	NaN
Best estimate	6.3	14.0
Precise	28.8	39.0

Arrival by ambulance

arrive_by_ambo = np.mean(data['S1ArriveByAmbulance'] == 'Yes') * 100
print (f'Percent arriving by ambulance: {arrive_by_ambo:0.1f}')

Percent arriving by ambulance: 81.4

# All admissions
ambulance_all = data.groupby('S1ArriveByAmbulance').mean()['thrombolysis_given']
ambulance_all.rename('All', inplace=True)

# Admissions within 4 hours of arrival
ambulance_4hr = data_4hr.groupby('S1ArriveByAmbulance').mean()['thrombolysis_given']
ambulance_4hr.rename('4hr', inplace=True)

# Combine
ambulance = pd.concat([ambulance_all, ambulance_4hr], axis=1) * 100
ambulance = ambulance.round(1)

ambulance

	All	4hr
S1ArriveByAmbulance
No	3.8	21.4
Yes	13.7	30.3

Thrombolysis use by comorbidities

Only examine effect of comorbidity on thrombolysis use of those arriving within 4 hours of known onset.

Analyse by comorbidity ‘Yes’ or ‘No’. Ignore ‘No but’.

comords = ['CongestiveHeartFailure', 'Hypertension', 'AtrialFibrillation', 
           'Diabetes', 'StrokeTIA', 'AFAntiplatelet', 'AFAnticoagulent', 
           'AFAnticoagulentVitK', 'AFAnticoagulentDOAC', 
           'AFAnticoagulentHeparin']

# Loop through comorbidities and get results
results = []
for comord in comords: 
    result = data_4hr.groupby(comord).mean()['thrombolysis_given']
    results.append({'No': result['No'] * 100, 
                    'Yes': result['Yes'] * 100})
    
comord_df = pd.DataFrame(results, index=comords)
comord_df['Ratio'] = comord_df['Yes'] / comord_df['No']
comord_df = comord_df.round(3)

comord_df

	No	Yes	Ratio
CongestiveHeartFailure	29.709	26.081	0.878
Hypertension	30.331	28.826	0.950
AtrialFibrillation	32.034	19.931	0.622
Diabetes	30.066	27.241	0.906
StrokeTIA	32.151	22.506	0.700
AFAntiplatelet	18.455	28.757	1.558
AFAnticoagulent	34.253	8.976	0.262
AFAnticoagulentVitK	31.009	11.942	0.385
AFAnticoagulentDOAC	32.254	5.677	0.176
AFAnticoagulentHeparin	30.068	7.576	0.252

# Set up figure
fig = plt.figure(figsize=(8,5))
ax = fig.add_subplot()
x = comord_df.index
y = comord_df['Ratio']

ax.barh(x, y)
ax.set_xlabel(
    'Ratio of thrombolysis use\n(co-morbidity present / comorbidity absent)')
ax.set_ylabel('Comorbidity')
ax.grid(axis='x')

plt.tight_layout()
plt.savefig('output/thrombolysis_use_by_comorbidity.jpg', dpi=300)
plt.show()

Reasons for not giving thrombolysis

Reasons given for not using thrombolysis (more than one may be ticked for each patient).

Only examine reasons for those arriving within 4 hours of known onset.

reasons = ['Haemorrhagic', 'TimeWindow', 'Comorbidity', 'Medication', 'Refusal',
           'Age', 'Improving', 'TooMildSevere', 'TimeUnknownWakeUp', 
           'OtherMedical']

# Restrict data to non-thrombolysed patients
mask = data_4hr['thrombolysis_given'] == False
no_thromolysis_4hr = data_4hr[mask]
# Convert Yes/No in reasons to boolean
reasons_boolean = no_thromolysis_4hr[reasons] == 'Yes'
# Get average of 'Yes' values as percentage
reasons = reasons_boolean.mean() * 100
reasons = reasons.round(1)
# Sort by value
reasons.sort_values(ascending=True, inplace=True)

reasons

Refusal               0.9
Age                   1.2
TimeWindow            7.1
Medication            7.7
Comorbidity           9.9
OtherMedical         15.2
TimeUnknownWakeUp    17.5
Improving            17.8
Haemorrhagic         21.7
TooMildSevere        22.0
dtype: float64

# Set up figure
fig = plt.figure(figsize=(8,5))
ax = fig.add_subplot()
x = reasons.index
y = reasons.values

ax.barh(x, y)
ax.set_xlabel('Presence of reason for no thrombolysis (%)')
ax.set_ylabel('Reason for no thrombolysis')
ax.grid(axis='x')

plt.tight_layout()
plt.savefig('output/thrombolysis_non_use_reasons.jpg', dpi=300)
plt.show()

Observations

• Disability before stroke: The higher the level of disability before stroke (as described by the modified Rankin Scale) the lower the use of thrombolysis. As a percentage of all out-of-hospital onset arrivals, thrombolysis use drops from 14.2% in those with no disability before stroke to 4.3% of those with a modified Rankin Scale of 5. As a percentage of out-of-hospital onset arrivals arriving within 4 hours of known stroke onset, thrombolysis use drops from 34.9% in those with no disability before stroke to 9.8% of those with a modified Rankin Scale of 5.

• Stroke severity: The use of thrombolysis varies very significantly with stroke severity (NIHSS) on arrival (Figure 2). Thrombolysis use is very low at extreme NIHSS, withy a plateau of use between NIHSS of approximately 6 to 25.

• Gender: There is a small association between gender on use of thrombolysis. 12.3% of all male arrivals receive thrombolysis, compared with 11.3% for females. Of those arriving within 4 hours of known stroke onset 30.8% of all male arrivals receive thrombolysis, compared with 28.2% for females.

• Ethnicity: There is some association between ethnicity on thrombolysis use in all arrivals, with black people having the lowest use of thrombolysis (10.2% vs 11.6% - 12.1% for white, asian and mixed race people), but little variation in use of thrombolysis in those arriving within 4 hours of known stroke onset, suggesting the cause of lower use of thrombolysis is likely to be the lower proportion of black people arriving at hospital within 4 hours of known stroke onset.

• Age: Thrombolysis use declines with age. For ages 25-55, thrombolysis use is about 15% of all arrivals and 35% of those arriving within 4 hours of known stroke onset. Above this age there is a decline in use, for example in the age band 85-89, use is about 10% of all arrivals and 25% of those arriving within 4 hours of known stroke onset

• Knowledge of time of onset: Knowledge of onset is split almost evenly (33% to 34% each) for ‘not known’, ‘best estimate’, and ‘precise’. Type of knowledge of stroke onset shoes a significant association with use of thrombolysis. Of those arriving within 4 hours of known onset, 39.0% receive thrombolysis if the time is recorded as being known precisely, compared with 14.0% receiving thrombolysis if the time is a best estimate.

• Arrival by ambulance: 8.4% of arrivals are by ambulance. There is a significant effect arrival model on use of thrombolysis. For all arrivals, use of thrombolysis is 13.7% for arrivals by ambulance, and 3.8% for other arrivals.

• Presence of co-morbidities: The presence or absence of co-co-morbidities can be a strong indicator of the use of thrombolysis. For example those on anticoagulant therapies receive thrombolysis less than those who do not, but those on anti-platelet therapies receive it more.

• Stated reasons for not giving thrombolysis:: Haemorrhagic stroke or a stroke that is too mild/sever are the most common causes (each present in 22% of those not treated). An improving condition is given as the reason for no treatment in 17% of non-treated patients. It is noteworthy that patient refusal is given as a reason for non-treatment in only 0.9% of untreated patients