Data on prognostic factors associated with 3-month and 1-year mortality from infective endocarditis

Abstract

This article describes supplementary tables and figures associated with the research paper entitled “Impact of referral bias on prognostic studies outcomes: insights from a population-based cohort study on infective endocarditis”. The aforementioned paper is a secondary analysis of data from the EI 2008 cohort on infective endocarditis and aimed at characterising referral bias. A total of 497 patients diagnosed with definite infective endocarditis between January 1^st and December 31^st 2008 were included in EI 2008. Data were collected from hospital medical records by trained clinical research assistants. Patients were divided into three groups: admitted to a tertiary hospital (group T), admitted to a non-tertiary hospital and referred secondarily to a tertiary hospital (group NTT) or admitted to a non-tertiary hospital and not referred (group NT). The pooled (NTT+T) group mimicked studies recruiting patients in tertiary hospitals only. Two different starting points were considered for follow up: date of first hospital admission and date of first admission to a tertiary hospital if any (hereinafter referred to as “referral time”). Referral bias is a type of selection bias which can occur due to recruitment of patients in tertiary hospitals only (excluding those who are admitted to non-tertiary hospitals and not referred to tertiary hospitals). This bias may impact the description of patients’ characteristics, survival estimates as well as prognostic factors identification. The six tables presented in this paper illustrate how patients’ selection (population-based sample [pooled (NT+NTT+T) group] versus recruitment in tertiary hospitals only [pooled (NTT+T) group]) might impact Hazards Ratios values for prognostic factors. Crude and adjusted Cox regression analyses were first performed to identify prognostic factors associated with 3-month and 1-year mortality in the whole sample using inclusion as the starting point. Analyses were then performed in the pooled (NTT+T) group first using inclusion as the starting point and finally using referral time as the starting point. Figures 1 to 3 illustrate how HR increase with time for covariates that were considered as time-varying covariates (covariate*time interaction).

Specifications Table

Subject	Medicine
Specific subject area	Epidemiology; Infectious Diseases
Type of data	Tables and figures
How data were acquired	Secondary analysis of data from the EI 2008 cohort on infective endocarditis
Data format	Analysed
Parameters for data collection	Patients presenting with a diagnosis of definite infective endocarditis (Duke criteria modified by Li) and admitted to a hospital between January 1st 2008 and December 31st 2008 in one of the seven participating French regions were included in the EI 2008 cohort. All the patients included in EI 2008 were considered for our analyses.
Description of data collection	Baseline and follow-up data were collected from hospital medical records by trained clinical research assistants and a standardized case report form was completed. Vital status was collected from hospitals medical records, general practitioners’ records or civil registry one year after inclusion in the EI 2008 cohort.
Data source location	Institution: CHRU Nancy City/Town/Region: Nancy Country: France Latitude and longitude for collected samples/data: 46.2276° N, 2.2137° E
Data accessibility	The datasets generated during and/or analysed during the current study are not publicly available due to restrictions pertaining to the French law, but are available from the corresponding author upon reasonable request.
Related research article	Collonnaz M, Erpelding M-L, Alla F, Goehringer F, Delahaye F, Iung B, et al. Impact of referral bias on prognostic studies outcomes: insights from a population-based cohort study on infective endocarditis. Annals of Epidemiology [Internet]. 2020 Sep; Available from: https://doi.org/10.1016/j.annepidem.2020.09.008[1]

Value of the Data

• •Studies recruiting patients in tertiary hospitals only are subject to referral bias. The results presented in the tables are part of a comprehensive analysis of the impact of this bias. They illustrate how referral bias can impact the assessment of the prognostic value of factors associated with 3-month and 1-year mortality from infective endocarditis.
• •Practitioners involved in infective endocarditis management as well as public health researchers and epidemiologists may benefit from these data. In addition, any researcher considering an observational study of a rare disease prone to be managed in non-tertiary hospitals at some point may benefit from these data.
• •These data may be useful for researchers aiming at characterising the impact of referral bias in studies on infective endocarditis or on other rare diseases. They can also be useful for future research on infective endocarditis as we highlighted the importance on a population-based recruitment of patients.

1. Data Description

The tables are supplementary data associated with the research paper entitled “Impact of referral bias on prognostic studies outcomes: insights from a population-based cohort study on infective endocarditis” [1]. The aforementioned paper is a secondary analysis of data from the EI 2008 cohort and aimed at characterising referral bias.

Tables 1 to 3 refer to the identification of prognostic factors associated with 3-month mortality from infective endocarditis (IE). Tables 4 to 6 refer to the identification of prognostic factors associated with to 1-year mortality.

Table 1.

Factors associated with 3-month mortality in the whole sample (pooled (NT+NTT+T) group) (Starting point=inclusion)

	Whole sample (N=460)
		Crude association			Adjusted association*
	n	HR	95% CI	p	HR	95% CI	p
Socio-demographic
Age				<0.001			<0.001
< 70	280	ref	-		ref	-
≥70	180	2.212	1.495-3.274		2.350	1.535-3.597
Sex				0.226
Female	117	ref	-
Male	343	0.771	0.505-1.175
Medical history
Charlson Comorbidity index				<0.001			0.009
<2	248	ref	-		ref	-
≥ 2	212	2.196	1.466-3.288		1.760	1.150-2.695
High blood pressure				0.006
No	242	ref	-
Yes	218	1.741	1.172-2.587
Injection drug use				0.242
No	435	ref	-
Yes	25	0.504	0.160-1.588
Underlying heart disease (HD)				0.491
No previously known HD	236	ref	-
Previously known HD without prosthetic valve	125	0.818	0.505-1.326
Prosthetic valve	99	1.148	0.714-1.847
Previous IE				0.119
No	430	ref	-
Yes	30	0.401	0.127-1.264
Cardiac implantable electronic device				0.464
No	399	ref	-
Yes	61	1.221	0.716-2.082
IE profile
Clinical characteristics
Suspected source of infection				0.071
Community	344	ref	-
Healthcare-related, acquired in hospital	103	1.510	0.982-2.323
Healthcare-related, not acquired in hospital	13	2.096	0.845-5.201
Left heart endocarditis				0.390
No	90	ref	-
Yes	370	1.257	0.747-2.115
Fever				0.586
No	65	ref	-
Yes	395	1.175	0.657-2.103
Heart failure				0.043
No	303	ref	-
Yes	157	1.499	1.013-2.218
Microbiological characteristics
Staphylococcal IE				<0.001			0.010
No	293	ref	-		ref	-
Yes	167	2.500	1.691-3.697		1.695	1.132-2.539
Echocardiographic characteristics
Vegetation				0.625
No vegetation	59	ref	-
≤15mm	204	0.856	0.459-1.597
>15mm	117	1.176	0.613-2.254
Unknown size of vegetation	80	0.997	0.488-2.034
Perforation				0.445
No	375	ref	-
Yes	85	0.812	0.476-1.385
IE complications
Cardiac abscess⁎⁎				0.048
No	361	ref	-
Yes	99	1.677	1.005-2.799
Septic shock⁎⁎				<0.001			<0.001
No	385	ref	-		ref	-
Yes	75	5.463	3.647-8.183		3.873	2.480-6.049
Cerebral haemorrhage⁎⁎				0.002
No	427	ref	-
Yes	33	2.584	1.414-4.724
Cerebral embolism⁎⁎				<0.001			<0.001
No	363	ref	-		ref	-
Yes	97	2.619	1.721-3.988		2.490	1.556-3.884
Vascular phenomena				0.083
No	242	ref	-
Yes	218	1.412	0.956-2.085
Immunologic phenomena				0.893
No	405	ref	-
Yes	55	0.960	0.525-1.752
Persistent sepsis despite treatment				0.085
No	410	ref	-
Yes	50	1.599	0.938-2.727
Serum creatinine level ≥180 µmol/L				<0.001			<0.001
No	327	ref	-		ref	-
Yes	133	4.857	3.254-7.248		3.160	2.081-4.797

Crude and adjusted Cox regressions were performed.

^⁎Stepwise selection, sle=0.20, sls=0.05

^⁎⁎Time dependent covariate

6 patients from group NTT were excluded from this analysis because they were not referred yet at 3 months.

Table 3.

Factors associated with 3-month mortality in patients who were managed in a tertiary hospital (Pooled (NTT+T) group) (Starting point=referral time)

	Pooled (NTT+T) group (N=410)
		Crude association			Adjusted association+			Adjusted association*
	n	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p
Socio-demographic
Age				0.001			<0.001			<0.001
< 70	261	ref	-		ref	-		ref	-
≥ 70	149	2.108	1.376-3.228		2.316	1.463-3.666		2.316	1.463-3.666
Sex
Female	100	ref	-	0.335
Male	310	0.793	0.494-1.271
Medical history
Charlson Comorbidity index				0.001			0.029			0.029
<2	230	ref	-		ref	-		ref	-
≥ 2	180	2.038	1.320-3.145		1.662	1.054-2.621		1.662	1.054-2.621
High blood pressure				0.005
No	225	ref	-
Yes	185	1.857	1.206-2.861
Injection drug use				0.359
No	386	ref	-
Yes	24	0.583	0.184-1.845
Underlying heart disease (HD)				0.399
No previously known HD	212	ref	-
Previously known HD without prosthetic valve	107	0.762	0.441-1.316
Prosthetic valve	91	1.168	0.701-1.945
Previous IE				0.196
No	382	ref	-
Yes	28	0.467	0.148-1.479
Cardiac implantable electronic device				0.599
No	3511	ref	-
Yes	59	1.166	0.658-2.069
IE profile
Clinical characteristics
Suspected source of infection				0.090
Community	304	ref	-
Healthcare-related, acquired in hospital	95	1.559	0.978-2.486
Healthcare-related, not acquired in hospital	11	2.082	0.754-5.746
Left heart endocarditis				0.354
No	82	ref	-
Yes	328	1.311	0.739-2.336
Fever				0.374
No	55	ref	-
Yes	355	1.368	0.686-2.730
Heart failure				0.085
No	266	ref	-
Yes	144	1.458	0.950-2.239
Microbiological characteristics
Staphylococcal IE				<0.001			0.007			0.007
No	263	ref	-		ref	-		ref	-
Yes	147	2.732	1.776-4.202		1.854	1.186-2.899		1.854	1.186-2.899
Echocardiographic characteristics
Vegetation				0.544
No vegetation	52	ref	-
≤15mm	176	0.968	0.477-1.963
>15mm	106	1.387	0.671-2.866
Unknown size of vegetation	76	1.036	0.465-2.306
Perforation				0.443
No	330	ref	-
Yes	80	0.799	0.451-1.418
IE complications
Cardiac abscess⁎⁎				0.060
No	315	ref	-
Yes	95	1. 669	0.979-2.846
Septic shock⁎⁎				<0.001			<0.001			<0.001
No	345	ref	.		ref	-		ref	-
Yes	65	5.057	3.239-7.894		3.462	2.119-5.657		3.462	2.119-5.657
Cerebral haemorrhage⁎⁎				0.001
No	381	ref	-
Yes	29	2.830	1.500-5.336
Cerebral embolism⁎⁎				<0.001			<0.001			<0.001
No	321	ref	-		ref	-		ref	-
Yes	89	2.579	1.638-4.061		2.518	1.544-4.105		2.518	1.544-4.105
Vascular phenomena				0.181
No	214	ref	-
Yes	196	1.338	0.874-2.051
Immunologic phenomena				0.674
No	359	ref	-
Yes	51	1.140	0.619-2.099
Persistent sepsis despite treatment				0.160
No	364	ref	.
Yes	46	1.528	0.846-2.758
Serum creatinine level ≥180 µmol/L				<0.001			<0.001			<0.001
No	291	ref	-		ref	-		ref	-
Yes	119	4.683	3.024-7.251		3.133	1.990-4.932		3.133	1.990-4.932

Crude and adjusted Cox regressions were performed.

^⁎Stepwise selection, sle=0.20, sls=0.05

⁺Covariates resulting from the selection process in the whole sample were forced into the model

^⁎⁎Time dependent covariate

Table 4.

Factors associated with 1-year mortality in the whole sample (pooled (NT+NTT+T) group) (Starting point=inclusion)

		Whole sample (N=466)
		Crude association			Adjusted association*
	n	HR	95% CI	p	HR	95% CI	p
Socio-demographic
Age				<0.001			<0.001
< 70	285	ref	-		ref	-
≥70	181	2.495	1.764-3.529		2.402	1.653-3.489
Sex				0.103
Female	118	ref	-
Male	348	0.736	0.509-1.064
Medical history
Charlson Comorbidity index⁎⁎⁎				<0.001			0.001
<2	252	ref	-		ref	-
≥ 2	214	1.011	1.006-1.016		1.008	1.003-1.013
High blood pressure⁎⁎⁎				<0.001			0.006
No	245	ref	-		ref	-
Yes	221	1.008	1.004-1.012		1.006	1.002-1.010
Injection drug use				0.079
No	440	ref	-
Yes	26	0.358	0.114-1.125
Underlying heart disease (HD)				0.806
No previously known HD	240	ref	-
Previously known HD without prosthetic valve	126	0.887	0.588-1.338
Prosthetic valve	100	1.029	0.667-1.587
Previous IE				0.078
No	436	ref	-
Yes	30	0.408	0.151-1.105
Cardiac implantable electronic device				0.276
No	403	ref	-
Yes	63	1.289	0.816-2.038
IE profile
Clinical characteristics
Suspected source of infection				0.033
Community	348	ref	-
Healthcare-related, acquired in hospital	105	1.540	1.055-2.246
Healthcare-related, not acquired in hospital	13	2.010	0.879-4.599
Left heart endocarditis				0.471
No	93	ref	-
Yes	373	1.177	0.756-1.831
Fever				0.559
No	66	ref	-
Yes	400	1.164	0.699-1.937
Heart failure⁎⁎⁎				0.001			0.027
No	307	ref	-		ref	-
Yes	159	1.006	1.002-1.009		1.004	1.000-1.007
Microbiological characteristics
Staphylococcal IE				<0.001			0.004
No	297	ref	-		ref	-
Yes	169	2.179	1.548-3.067		1.679	1.177-2.395
Echocardiographic characteristics
Vegetation				0.885
No vegetation	59	ref	-
≤15mm	207	1.051	0.596-1.855
>15mm	119	1.221	0.669-2.231
Unknown size of vegetation	81	1.099	0.573-2.106
Perforation				0.143			0.020
No	381	ref	-		ref	-
Yes	85	0.690	0.420-1.134		0.545	0.327-0.909
IE complications
Cardiac abscess⁎⁎				0.283
No	367	ref	-
Yes	99	1.299	0.806-2.096
Septic shock⁎⁎				<0.001			<0.001
No	390	ref	-		ref	-
Yes	76	4.129	2.840-6.003		3.360	2.211-5.104
Cerebral haemorrhage⁎⁎				0.015
No	433	ref	-
Yes	33	2.042	1.151-3.622
Cerebral embolism⁎⁎				<0.001			<0.001
No	369	ref	-		ref	-
Yes	97	2.256	1.545-3.293		2.324	1.556-3.469
Vascular phenomena⁎⁎⁎				0.182
No	246	ref	-
Yes	220	0.998	0.094-1.001
Immunologic phenomena				0.272
No	411	ref	-
Yes	55	0.717	0.396-1.298
Persistent sepsis despite treatment				0.091
No	416	ref	-
Yes	50	1.521	0.936-2.474
Serum creatinine level ≥180 µmol/L				<0.001			<0.001
No	333	ref	-		ref	-
Yes	133	3.996	2.832-5.638		2.780	1.944-3.977

Crude and adjusted Cox regressions were performed.

^⁎Stepwise selection, sle=0.20, sls=0.05

^⁎⁎Time dependent covariate

^⁎⁎⁎Time-varying covariate (interaction with time)

Table 6.

Factors associated with 1-year mortality in patients who were managed in a tertiary hospital (Pooled (NTT+T) group) (Starting point=referral time)

	Pooled (NTT+T) group (N=410)
		Crude association			Adjusted association+			Adjusted association*
	n	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p
Socio-demographic
Age				<0.001			<0.001			<0.001
< 70	261	ref	-		ref	-		ref	-
≥ 70	149	2.343	1.586-3.462		2.310	1.520-3.511		2.317	1.526-3.518
Sex				0.261
Female	100	ref	-
Male	310	0.781	0.508-1.201
Medical history
Charlson Comorbidity index⁎⁎⁎				0.003			0.036			0.022
<2	230	ref	-		ref	-		ref	-
≥ 2	180	1.008	1.003-1.013		1.005	1.000-1.010		1.006	1.001-1.011
High blood pressure⁎⁎⁎				0.001			0.005			0.004
No	225	ref	-		ref	-		ref	-
Yes	185	1.011	1.005-1.018		1.009	1.003-1.016		1.009	1.003-1.016
Injection drug use				0.201
No	386	ref	-
Yes	24	0.473	0.150-1.491
Underlying heart disease (HD)				0.453
No previously known HD	212	ref	-
Previously known HD without prosthetic valve	107	0.823	0.505-1.343
Prosthetic valve	91	1.180	0.738-1.887
Previous IE				0.193
No	382	ref	-
Yes	28	0.515	0.190-1.400
Cardiac implantable electronic device				0.507
No	351	ref	-
Yes	59	1.193	0.709-2.008
IE profile
Clinical characteristics
Suspected source of infection				0.039
Community	304	ref	-
Healthcare-related, acquired in hospital	95	1.581	1.031-2.422
Healthcare-related, not acquired in hospital	11	2.239	0.902-5.558
Left heart endocarditis				0.324
No	82	ref	-
Yes	328	1.300	0.772-2.188
Fever				0.520
No	55	ref	-
Yes	355	1.219	0.667-2.226
Heart failure⁎⁎⁎				0.006			0.081
No	266	ref	-		ref	-
Yes	144	1.006	1.002-1.011		1.004	1.000-1.008
Microbiological characteristics
Staphylococcal IE				<0.001			0.020			0.017
No	263	ref	-		ref	-		ref	-
Yes	147	2.225	1.508-3.283		1.618	1.079-2.426		1.633	1.092-2.443
Echocardiographic characteristics
Vegetation				0.729
No vegetation	52	ref	-
≤15mm	176	1.153	0.594-2.235
>15mm	106	1.413	0.708-2.820
Unknown size of vegetation	76	1.131	0.534-2.394
Perforation				0.118			0.038
No	330	ref	-		ref	-
Yes	80	0.638	0.363-1.121		0.543	0.305-0.966
IE complications
Cardiac abscess⁎⁎				0.132
No	315	ref	-
Yes	95	1.469	0.891-2.422
Septic shock⁎⁎				<0.001			<0.001			<0.001
No	345	ref	-		ref	-		ref	-
Yes	65	4.086	2.679-6.233		2.847	1.754-4.621		2.766	1.721-4.445
Cerebral haemorrhage⁎⁎				0.003
No	381	ref	-
Yes	29	2.494	1.364-4.558
Cerebral embolism⁎⁎				<0.001			<0.001			<0.001
No	321	ref	-		ref	-		ref	-
Yes	89	2.336	1.536-3.555		2.499	1.581-3.949		2.551	1.615-4.029
Vascular phenomena⁎⁎⁎				0.168
No	214	ref	-
Yes	196	0.997	0.993-1.001
Immunologic phenomena				0.810
No	359	ref	-
Yes	51	0.929	0.509-1.697
Persistent sepsis despite treatment				0.171
No	364	ref	-
Yes	46	1.466	0.848-2.536
Serum creatinine level ≥180 µmol/L				<0.001			<0.001			<0.001
No	291	ref	-		ref	-		ref	-
Yes	119	4.113	2.776-6.095		3.085	2.053-4.634		3.055	2.031-4.594

Crude and adjusted Cox regressions were performed.

^⁎Stepwise selection, sle=0.20, sls=0.05

⁺Covariates resulting from the selection process in the whole sample were forced into the model

^⁎⁎Time dependent covariate

^⁎⁎⁎Time-varying covariate (interaction with time)

Tables 1 and 4 present absolute frequency, crude and adjusted Hazards Ratios (HR) when using the whole sample of patients (pooled (NT+NTT+T) group) and using the date of inclusion (date of first admission to hospital for IE) as the starting point for analyses.

Tables 2 and 5 present absolute frequency, crude and adjusted HRs obtained when performing analyses on a sample of patients recruited in tertiary hospitals only (pooled (NTT+T) group) using date of inclusion as the starting point.

Table 2.

Factors associated with 3-month mortality in patients who were managed in a tertiary hospital (Pooled (NTT+T) group) (Starting point=inclusion)

	Pooled (NTT+T) group (N=404)
		Crude association			Adjusted association+			Adjusted association*
	n	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p
Socio-demographic
Age				0.001			<0.001			<0.001
< 70	256	ref	-		ref	-		ref	-
≥ 70	148	2.085	1.361-3.193		2.368	1.494-3.755		2.368	1.494-3.755
Sex				0.391
Female	99	ref	-
Male	305	0.813	0.507-1.304
Medical history
Charlson Comorbidity index				0.002			0.020			0.020
<2	226	ref	-		ref	-		ref	-
≥ 2	178	2.011	1.303-3.105		1.712	1.087-2.699		1.712	1.087-2.699
High blood pressure				0.005
No	222	ref	-
Yes	182	1.851	1.202-2.852
Injection drug use				0.362
No	381	ref	-
Yes	23	0.585	0.185-1.851
Underlying heart disease (HD)				0.379
No previously known HD	208	ref	-
Previously known HD without prosthetic valve	106	0.747	0.433-1.291
Prosthetic valve	90	1.157	0.695-1.926
Previous IE				0.182
No	376	ref	-
Yes	28	0.457	0.144-1.445
Cardiac implantable electronic device				0.531
No	347	ref	-
Yes	57	1.201	0.677-2.130
IE profile
Clinical characteristics
Suspected source of infection				0.100
Community	300	ref	-
Healthcare-related, acquired in hospital	93	1.541	0.967-2.457
Healthcare-related, not acquired in hospital	11	2.061	0.747-5.687
Left heart endocarditis				0.384
No	79	ref	-
Yes	325	1.290	0.727-2.288
Fever				0.361
No	54	ref	-
Yes	350	1.380	0.692-2.754
Heart failure				0.088
No	262	ref	-
Yes	142	1.452	0.946-2.229
Microbiological characteristics
Staphylococcal IE				<0.001			0.006			0.006
No	259	ref	-		ref	-		ref	-
Yes	145	2.733	1.777-4.204		1.863	1.191-2.914		1.863	1.191-2.914
Echocardiographic characteristics
Vegetation				0.495
No vegetation	52	ref	-
≤15mm	173	0.965	0.476-1.958
>15mm	104	1.414	0.685-2.922
Unknown size of vegetation	75	1.045	0.470-2.326
Perforation				0.396
No	324	ref	-
Yes	80	0.780	0.440-1.384
IE complications
Cardiac abscess⁎⁎				0.022
No	309	ref	-
Yes	95	1.869	1.094-3.193
Septic shock⁎⁎				<0.001			<0.001			<0.001
No	340	ref	-		ref	-		ref	-
Yes	64	5.288	3.387-8.254		3.622	2.209-5.937		3.622	2.209-5.937
Cerebral haemorrhage⁎⁎				0.001
No	375	ref	-
Yes	29	2.941	1.559-5.548
Cerebral embolism⁎⁎				<0.001			<0.001			<0.001
No	315	ref	-		ref	-		ref	-
Yes	89	2.618	1.662-4.125		2.563	1.567-4.190		2.563	1.567-4.190
Vascular phenomena				0.193
No	210	ref	-
Yes	194	1.328	0.866-2.034
Immunologic phenomena				0.711
No	353	ref	-
Yes	51	1.123	0.610-2.067
Persistent sepsis despite treatment				0.219
No	358	ref	-
Yes	46	1.449	0.803-2.615
Serum creatinine level ≥180 µmol/L				<0.001			<0.001			<0.001
No	285	ref	-		ref	-		ref	-
Yes	119	4.533	2.928-7.019		3.048	1.935-4.799		3.048	1.935-4.799

Crude and adjusted Cox regressions were performed.

^⁎Stepwise selection, sle=0.20, sls=0.05

⁺Covariates resulting from the selection process in the whole sample were forced into the model

^⁎⁎Time dependent covariate

6 patients from group NTT were excluded from this analysis because they were not referred yet at 3 months.

Table 5.

Factors associated with 1-year mortality in patients who were managed in a tertiary hospital (Pooled (NTT+T) group) (Starting point=inclusion)

	Pooled (NTT+T) group (N=410)
		Crude association			Adjusted association+			Adjusted association*
	n	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p
Socio-demographic
Age				<0.001			<0.001			<0.001
< 70	261	ref	-		ref	-		ref	-
≥ 70	149	2.253	1.543-3.290		2.193	1.460-3.293		2.193	1.460-3.293
Sex				0.271
Female	100	ref	-
Male	310	0.791	0.521-1.200
Medical history
Charlson Comorbidity index⁎⁎⁎							0.005			0.005
<2	230	ref	-	<0.001	ref	-		ref	-
≥ 2	180	1.009	1.005-1.014		1.007	1.002-1.012		1.007	1.002-1.012
High blood pressure⁎⁎⁎							0.007			0.007
No	225	ref	-	<0.001	ref	-		ref	-
Yes	185	1.008	1.004-1.013		1.006	1.002-1.011		1.006	1.002-1.011
Injection drug use				0.144
No	386	ref	-
Yes	24	0.425	0.135-1.340
Underlying heart disease (HD)				0.664
No previously known HD	212	ref	-
Previously known HD without prosthetic valve	107	0.862	0.541-1.373
Prosthetic valve	91	1.106	0.694-1.760
Previous IE				0.149
No	382	ref	-
Yes	28	0.479	0.177-1.302
Cardiac implantable electronic device				0.341
No	351	ref	-
Yes	59	1.272	0.775-2.088
IE profile
Clinical characteristics
Suspected source of infection				0.075
Community	304	ref	-
Healthcare-related, acquired in hospital	95	1.484	0.978-2.253
Healthcare-related, not acquired in hospital	11	2.071	0.836-5.129
Left heart endocarditis				0.268
No	82	ref	-
Yes	328	1.331	0.802-2.208
Fever				0.362
No	55	ref	-
Yes	355	1.322	0.725-2.408
Heart failure⁎⁎⁎				<0.001			0.017			0.017
No	266	ref	-		ref	-		ref	-
Yes	144	1.007	1.003-1.011		1.005	1.001-1.009		1.005	1.001-1.009
Microbiological characteristics
Staphylococcal IE				<0.001			0.015			0.015
No	263	ref	-		ref	-		ref	-
Yes	147	2.153	1.476-3.141		1.634	1.102-2.423		1.634	1.102-2.423
Echocardiographic characteristics
Vegetation				0.607
No vegetation	52	ref	-
≤15mm	176	1.221	0.633-2.358
>15mm	106	1.523	0.768-3.022
Unknown size of vegetation	76	1.191	0.567-2.503
Perforation				0.178			0.047			0.047
No	330	ref	-		ref	-		ref	-
Yes	80	0.694	0.408-1.181		0.574	0.332-0.992		0.574	0.332-0.992
IE complications
Cardiac abscess⁎⁎				0.116
No	315	ref	-
Yes	95	1.491	0.906-2.455
Septic shock⁎⁎				<0.001			<0.001			<0.001
No	345	ref	-		ref	-		ref	-
Yes	65	3.962	2.610-6.013		2.991	1.874-4.773		2.991	1.874-4.773
Cerebral haemorrhage⁎⁎				0.005
No	381	ref	-
Yes	29	2.386	1.308-4.352
Cerebral embolism⁎⁎				<0.001			<0.001			<0.001
No	321	ref	-		ref	-		ref	-
Yes	89	2.184	1.443-3.307		2.266	1.452-3.535		2.266	1.452-3.535
Vascular phenomena⁎⁎⁎				0.105
No	214	ref	-
Yes	196	0.997	0.993-1.001
Immunologic phenomena				0.624
No	359	ref	-
Yes	51	0.861	0.472-1.568
Persistent sepsis despite treatment				0.166
No	364	ref	-
Yes	46	1.455	0.856-2.475
Serum creatinine level ≥180 µmol/L				<0.001			<0.001			<0.001
No	291	ref	-		ref	-		ref	-
Yes	119	3.942	2.693-5.770		2.882	1.941-4.278		2.882	1.941-4.278

Crude and adjusted Cox regressions were performed.

^⁎Stepwise selection, sle=0.20, sls=0.05

⁺Covariates resulting from the selection process in the whole sample were forced into the model

^⁎⁎Time dependent covariate

^⁎⁎⁎Time-varying covariate (interaction with time)

Tables 3 and 6 present absolute frequency, crude and adjusted HRs obtained when recruiting patients in tertiary hospitals only (pooled (NTT+T) group) using referral time (the date of first admission to a tertiary hospital if any) as the starting point.

A total of six prognostic factors were associated with 3-month mortality (age ≥70, Charlson comorbidity index ≥2, Staphylococcal IE, septic shock, cerebral embolism, and serum creatinine level ≥18 μmol/L). The prognostic factors did not differ across groups; however, the values of HRs associated with these prognostic factors were influenced by sample and starting point selection.

When using the date of inclusion as the starting point, a total of nine prognostic factors were associated with 1-year mortality in the whole sample and in the pooled (NTT+T) group (age ≥70, Charlson comorbidity index ≥2*time, high blood pressure*time, heart failure*time, Staphylococcal IE, valve perforation, septic shock, cerebral embolism, and serum creatinine level ≥180 μmol/L). When using referral time as the starting point, heart failure and valve perforation were no longer identified as prognostic factors.

Figs. 1 to 3 represent the evolution of Hazard Ratios with time (in days) for the three covariates (Charlson comorbidity index ≥2, high blood pressure and heart failure) that did not meet the proportional hazard assumption and were included in 1-year survival models as time-varying covariates (considering covariate*time interaction). Figs. 1 to 3 show how HR for these covariates increased with time. For example, for Charlson comorbidity index ≥2, the risk of death from IE in the whole sample was multiplied by 1.27 [1.09–1.47] after one month and by 1.61 [1.20–2.17] after two months. Noteworthy, 95% confidence intervals width also increased with time, showing a loss of precision in HR estimates over time.

Fig. 2.

Evolution over time of Hazard Ratios for the association between high blood pressure and one-year mortality.

Fig. 1.

Evolution over time of Hazard Ratios for the association between Charlson comorbidity index ≥2 and one-year mortality.

Fig. 3.

Evolution over time of Hazard Ratios for the association between heart failure and one-year mortality.

2. Experimental Design, Materials and Methods

2.1. The EI 2008 cohort

2.1.1. Background

We used data from EI 2008, a one-year prospective, population based, cohort study of patients with IE [2]. Inclusion criteria in EI 2008 were: being over 18 years old, living in one of the seven participating French administrative areas (greater Paris, Lorraine, Rhône-Alpes, Franche-Comté, Marne, Ille-et-Villaine and Languedoc- Roussillon), and being admitted to hospital for IE between January 1st 2008 and December 31^st 2008. Diagnosis of IE was adjudicated by a team of infectious diseases professionals. Definite cases of IE (Duke criteria modified by Li [3]) were included in the cohort. Patients were followed during one year after inclusion. Baseline and follow-up data were collected on a standardised case report form by trained clinical research assistants. Information on patients’ characteristics, IE profiles, treatment and complications were retrieved from hospital medical records. Vital status was assessed from hospital medical records, general practitioners’ records or civil registry office 1 year after inclusion, and date of death was collected when appropriate.

All the patients enrolled in the EI 2008 cohort (497 patients with a definite diagnosis of IE) were considered for our analyses.

2.1.2. Data collection

Patients were divided into three groups: patients admitted to a tertiary hospital (group T), patients admitted to a non-tertiary hospital and secondarily referred to a tertiary hospital (group NTT), and patients admitted to a non-tertiary hospital and not secondarily referred to a tertiary hospital (group NT).

Baseline data consisted of sociodemographic characteristics (age ≥70, sex) and medical history data (Charlson comorbidity index ≥2, high blood pressure, injection drug use, underlying heart disease, previous IE and implantable cardiac device). Data on IE profile and IE complications were collected during follow-up in hospital stay. IE profile data included clinical characteristics (suspected source of infection, left heart endocarditis, fever and heart failure), microbiological characteristics (Staphylococcal IE), and echocardiographic characteristics (vegetation and perforation). IE complications data consisted of cardiac abscess, septic shock, cerebral haemorrhage, cerebral embolism, vascular phenomena, immunologic phenomena, persistent sepsis despite treatment, and serum creatinine level ≥180 µmol/L.

2.2. Secondary analysis of EI 2008 data

2.2.1. Aim of the study

We used data from the EI 2008 cohort to provide a comprehensive characterisation of referral bias [1]. Referral bias is a type of selection bias that can occur in studies recruiting patients in tertiary hospitals only (mixing patients admitted directly to tertiary hospitals and those referred secondarily to tertiary hospitals, but excluding patients admitted to non-tertiary hospitals and not referred). Studies on rare diseases such as infective endocarditis are particularly prone to referral bias [4], [5], [6], a bias which may threaten the validity of prognostic studies’ results [7].

2.2.2. Data analyses

The whole sample (pooled (NT+NTT+T) group) represented a population-based recruitment of patients with IE. The pooled (NTT+T) group was used to mimic prognostic studies recruiting patients in tertiary hospitals.

Two different starting points were considered for the follow-up: inclusion (corresponding to the date of first hospital admission) and referral time (corresponding to the date of first admission to a tertiary hospital if any, i.e. used for groups T and NTT only to mimic prognostic studies based on patients recruited in tertiary hospitals only). Survival time was calculated from the starting point to the date of death or to the date of last follow-up.

All variables mentioned in data collection were evaluated for their potential prognostic impact. Patients with missing data for at least one of these potential prognostic factors were excluded from analyses (after checking that the characteristics of excluded patients did not differ significantly from those of included patients). One patient (patient number 560) was excluded from analyses due to a negative delay between hospital admission and septic shock (covariate introduced in Cox analyses as a time-dependent covariate). As a result, a total of 466 patients with infective endocarditis were included in the Cox analyses (274 in group T (58,8%), 136 in group NTT (29,2%) and 56 in group NT (12,0%)).

Data presented in tables 1 to 6 were obtained through crude and adjusted Cox modelling of 3-month and 1-year survival from IE. For both 3-months and 1-year survival, analyses were performed in five steps:

• 1/ In the pooled (NT+NTT+T) group, i.e. a population-based sample, using inclusion as the starting point, a stepwise selection (with a significance level for entry (sle) set at 0.20 and a significance level for staying in the model (sls) set at 0.05) was performed to identify prognostic factors

• 2/ In the pooled (NTT+T), using inclusion as the starting point, covariates identified as prognostic factors in Step 1 were forced into the multivariable model to identify an eventual difference in significance or in the magnitude of HRs

• 3/ In the pooled (NTT+T) group, using inclusion as the starting point, a stepwise selection (sle=0.20 and sls=0.05) was performed to naively identify prognostic factors

• 4/ In the pooled (NTT+T) group, using referral time as the starting point, covariates identified as prognostic factors in Step 1 were forced into the multivariable model to identify an eventual difference in the magnitude or significance of their effect

• 5/ In the pooled (NTT+T) group, using referral time as the starting point, a stepwise selection (sle=0.20 and sls=0.05) was performed to identify prognostic factors using a design prone to referral bias.

Cox models assumptions (Log-linearity assumption and proportional hazard assumption) were checked. In 1-year Cox analyses, high blood pressure, Charlson comorbidity index, heart failure, and vascular phenomena did not meet the proportional hazard assumption and were considered as time-varying covariates (covariate*time interaction). IE complications occurring during the hospital stay (septic shock, cardiac abscess, cerebral embolism, and cerebral haemorrhage) were introduced in the models as time-dependent covariates. Statistical analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, NC) software.

Ethics Statement

EI 2008 was conducted in accordance with the Declaration of Helsinki. Patients were informed of the study but their written individual consent was not required. EI 2008 was authorized by the Commission Nationale de l'Informatique et des Libertés (CNIL-DR-2010-219) and registered in ClinicalTrials.gov (NCT03295045).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships which have, or could be perceived to have, influenced the work reported in this article.