Abstract
Purpose
To assess the risk of acute infections in patients with first ever diagnosed depression compared to patients with no diagnosed depression in a primary-care database.
Methods
We conducted a cohort study using the UK CPRD GOLD database (2000–2019). We identified patients aged 18 years or older with a recorded Read code for depression (cohort entry date) and compared them to patients with no Read codes for depression using risk set sampling. Comparison groups were frequency-matched on age and sex, and comparison patients were required to have ≥ 1 general practitioner (GP) contact within 14 days before cohort entry. The primary outcome was a composite of outpatient diagnosed acute infections, including respiratory, gastrointestinal, urogenital infections and septicemia) within the two-years after cohort entry. We applied propensity score fine stratification and estimated incidence rates and IR ratios (IRR) using negative binomial regression.
Results
In a weighted population of 285,922 patients with diagnosed depression and 285,921 comparison patients, the IR of acute infections was 97.3/1000 person-years (py) in patients with and 83.7/1000 py in patients with no diagnosed depression. The weighted IRR of acute infection was 1.18 (95% CI 1.16–1.20) comparing those with and with no diagnosed depression. Excluding patients with baseline comorbidities yielded an IRR even closer to the null: 1.07 (95% CI, 1.04–1.09).
Conclusions
Our results suggest that patients with diagnosed depression are not at a meaningfully increased risk of acute infections compared to patients with no diagnosed depression. Slightly increased overall relative risks of infections can be explained by residual differences in health care utilization and by the severity of comorbidities.
Supplementary Information
The online version contains supplementary material available at 10.1007/s00127-024-02784-1.
Keywords: Depression, infection risk, acid Sphingomyelinase, CPRD
Introduction
The lifetime prevalence of major depressive episodes is estimated to be between 15% and 18% worldwide [1–4].
In vitro studies have indicated that alterations in the immune system are present in patients with depression [5–7]. However, the extent to which these changes result in an increased risk of acute infections in the general population is poorly understood [8, 9].
Two previous observational studies reported an up to 60% increased risk of infection in patients with depression compared to patients without depression [10, 11]. One study was conducted using Danish health registry data, which are representative of the overall Danish population (1995–2012). The second study was a survey among US college students (2004). However, neither study controlled for important potential confounders such as BMI, comorbidities, or health care utilization (surveillance bias). Furthermore, the most frequently recorded infectious disease in the Danish study was hepatic infection (even before respiratory infections) [11–13], which is likely not representative of the infection burden of the overall general population. Thus, the robustness and representativeness of these prior results is questionable. Overall, evidence of the association between depression and the risk of infection in the general population is scarce.
We conducted an observational cohort study using a UK-based primary care database to compare the risk of acute infections between adults with and with no diagnosed depression.
Patients and methods
Study design and data source
We conducted a cohort study using data from the UK Clinical Practice Research Datalink (CPRD) GOLD, a primary care database that contains de-identified health data on some 14 million patients recorded by general practitioners (GPs) [14–16]. CPRD GOLD has been described extensively elsewhere [14–16] but briefly, it captures information on patient demographics, lifestyle factors, diagnoses, medication prescriptions, laboratory test results, and referrals to secondary care.
Study population
We conducted a cohort study comparing patients with an incident diagnosis of depression between January 1, 2000, and December 31, 2019, defined as having a first ever recorded Read code for depression (Read-code list available in supplement) to patients with no recorded Read code for depression. All patients were aged 18 years or older.
All patients were required to have at least 2 years of recorded history in CPRD GOLD before the cohort entry date (date of first depression diagnosis). Patients who had a prescription for an antidepressant at any time before the cohort entry date were excluded. If a code for suicidal ideation or suicide attempt was recorded within one year prior to the first diagnosis of depression, the cohort entry date was set to the first record of suicidal ideation (N = 3984, 0.6% of all patients with diagnosed depression) [17]. For each exposed patient, we selected one comparison patient with the same cohort entry date, but without a recorded Read code for depression at any time prior to the cohort entry date of the corresponding exposed patients. Comparison patients were matched to exposed patients on year of birth, sex, and post code and were required to have at least one GP visit within the last 14 days before cohort entry date to ensure that the patient was actively enrolled with his or her GP at that time.
We further excluded patients with a recorded diagnosis of alcoholism or HIV at any time before the cohort entry date, as well as those with a diagnosis of cancer / ongoing chemotherapy or with use of immunosuppressive drugs within two years before the cohort entry date.
Study outcomes
The primary outcome was a composite of four types of infections identified using Read diagnosis codes for (1) acute respiratory, (2) genitourinary, or (3) gastrointestinal infections, as well as (4) septicaemia. All recorded infections during follow-up were captured; thus, a patient could contribute more than one outcome event. A time lag of at least 14 days was required between two records of infection of the same type to be counted as separate events. Code lists are available in the supplement. In a secondary analysis, we evaluated each type of infection separately, not censoring for occurrence of another type of infection.
Additionally, we identified all patients in the study population with HES linkage (approx. 50%). In this subset, we identified hospital-diagnosed infections (composite outcome of the same four types of acute infections included in the main analyses) using ICD-10 codes (ICD-10 codes are listed in the supplement).
Follow-up
We accumulated person time for patients with and with no diagnosed depression (exposed or non-exposed) starting on the first day after the cohort entry date until censoring. Patients were censored on the date the first of the following occurred: they became a case, an exclusion criterion was recoded, death, end of data recording, the end of the study period, or after a maximum of 630 days (2 years) of follow up. Patients with diagnosed depression were censored after 12 months (on day 366) if there was no further record for either a prescription of an antidepressant or a repeated diagnosis of depression. In addition, we censored comparison patients when a diagnosis of depression or a prescription for an antidepressant was recorded during follow-up.
Patient characteristics and covariables
We evaluated 32 baseline covariables, including demographics, lifestyle factors, and comorbidities associated with health care utilization at any time prior to the cohort entry date. We further assessed a history of cancer > 2 years prior to the cohort entry date, or a recorded diagnosis for pneumonia within 180 days prior to cohort entry date. We further captured prescriptions for immunosuppressive medication recorded within 180 days prior to the cohort entry date, as well as the number of GP contacts (health care utilization) within 1 year before the cohort entry date. All covariables are listed in Table 1.
Table 1.
Characteristics of patients with and without first ever diagnosed depression at cohort entry before and after PS-weighting
| Before PS-weighting No. (%) |
After PS-weighting No. (%) |
|||
|---|---|---|---|---|
|
Diagnosed
depression |
Without diagnosed depression | Diagnosed depression | Without diagnosed depression | |
| No. of patients | 285 923 | 285 923 | 285 922 | 285 921 |
| Age, mean (SD), y | 38.9 (17.5) | 38.9 (17.5) | 38.9 (17.5) | 39.3 (17.5) |
| Women | 177 438 (62.1) | 177 438 (62.1) | 177 438 (62.1) | 178 089 (62.3) |
| No. of GP visits (365 d prior), mean (SD) | 8.3 (8.4) | 9.4 (8.7) | 8.3 (8.4) | 8.5 (7.8) |
| BMI | ||||
| 12.0-18.4 | 7 668 (2.7) | 6 896 (2.4) | 7 668 (2.7) | 7 627 (2.7) |
| 18.5–24.9 | 93 362 (32.7) | 97 825 (34.2) | 93 362 (32.7) | 92 318 (32.3) |
| 25.0–29.9 | 62 780 (22) | 66 741 (23.3) | 62 780 (22) | 63 310 (22.1) |
| 30.0–60.0 | 47 443 (16.6) | 47 173 (16.5) | 47 443 (16.6) | 48 642 (17) |
| Unknown | 74 670 (26.1) | 67 288 (23.5) | 74 669 (26.1) | 74 024 (25.9) |
| Smoking status | ||||
| Non-smoker | 120 950 (42.3) | 153 100 (53.5) | 120 949 (42.3) | 118 675 (41.5) |
| Current smoker | 76,753 (26.8) | 53,181 (18.6) | 76 753 (26.8) | 78 659 (27.5) |
| Past smoker | 55 152 (19.3) | 54 536 (19.1) | 55 152 (19.3) | 55 429 (19.4) |
| Unknown | 33 068 (11.6) | 25 106 (8.8) | 33 068 (11.6) | 33 158 (11.6) |
| Alcohol consuption | ||||
| Low risk consumption ( < = 14 units/week) | 187 612 (65.6) | 193 630 (67.7) | 187 612 (65.6) | 188 315 (65.9) |
| High risk consumption (14 + units/week) | 15 481 (5.4) | 14 458 (5.1) | 15 481 (5.4) | 15 497 (5.4) |
| Unknown | 82 830 (29) | 77 835 (27.2) | 82 829 (29) | 82 110 (28.7) |
| Comedication (180 d prior) | ||||
| Opioid | 13 936 (4.9) | 13 854 (4.8) | 13 936 (4.9) | 14 880 (5.2) |
| Antipsychotics | 6 266 (2.2) | 6 304 (2.2) | 6 266 (2.2) | 6 574 (2.3) |
| Agents acting on RAS | 19 927 (7) | 27 795 (9.7) | 19 927 (7) | 21 464 (7.5) |
| Beta-blocker | 16 314 (5.7) | 19 233 (6.7) | 16 314 (5.7) | 17 395 (6.1) |
| Calcium channel blockers | 11 850 (4.1) | 16 011 (5.6) | 11 850 (4.1) | 12 852 (4.5) |
| Diuretics | 17 621 (6.2) | 22 960 (8) | 17 621 (6.2) | 19 167 (6.7) |
| Statins | 18 596 (6.5) | 23 838 (8.3) | 18 596 (6.5) | 20 245 (7.1) |
| Thrombocyte aggregation inhibitors | 15 547 (5.4) | 17 514 (6.1) | 15 547 (5.4) | 16 782 (5.9) |
| Anticoagulants | 2 802 (1) | 4 607 (1.6) | 2 802 (1) | 3 271 (1.1) |
| Proton pump inhibitors | 21 359 (7.5) | 24 683 (8.6) | 21 359 (7.5) | 22 531 (7.9) |
| Systemic corticosteroids | 8 939 (3.1) | 12 031 (4.2) | 8 939 (3.1) | 9 504 (3.3) |
| Immunosuppressant | 502 (0.2) | 896 (0.3) | 502 (0.2) | 599 (0.2) |
| Comorbidities | ||||
| Asthma | 51 150 (17.9) | 56 603 (19.8) | 51 150 (17.9) | 50 958 (17.8) |
| COPD | 3 907 (1.4) | 3 432 (1.2) | 3 907 (1.4) | 4 098 (1.4) |
| Arrhythmia | 4 404 (1.5) | 4 791 (1.7) | 4 404 (1.5) | 4 574 (1.6) |
| Congestive heart failure | 2 857 (1) | 2 858 (1) | 2 857 (1) | 3 026 (1.1) |
| Ischemic heart disease | 9 149 (3.2) | 9 084 (3.2) | 9 149 (3.2) | 9 763 (3.4) |
| Myocardial infarction | 4 132 (1.4) | 4 357 (1.5) | 4 132 (1.4) | 4 464 (1.6) |
| Hyperlipidaemia | 13 268 (4.6) | 15 598 (5.5) | 13 268 (4.6) | 13 985 (4.9) |
| Diabetes mellitus | 12 026 (4.2) | 17 710 (6.2) | 12 025 (4.2) | 13 141 (4.6) |
| DVT / PE | 833 (0.3) | 1 027 (0.4) | 833 (0.3) | 866 (0.3) |
| Ulcer | 5 004 (1.8) | 4 430 (1.5) | 5 004 (1.8) | 5 162 (1.8) |
| Diagnosed renal diseases | 5 970 (2.1) | 6 289 (2.2) | 5 970 (2.1) | 6 146 (2.1) |
| Hypothyroidism | 9 874 (3.5) | 13 796 (4.8) | 9 874 (3.5) | 10 455 (3.7) |
| History of cancer (non-active, last code ≥ 730 days before CED) | 7 580 (2.7) | 8 227 (2.9) | 7 580 (2.7) | 7 818 (2.7) |
| Pneumonia (any time up to 180 days prior) | 5 907 (2.1) | 5 726 (2) | 5 907 (2.1) | 5 947 (2.1) |
| C-statistic | 0.60 | 0.50 | ||
Abbreviations: BMI, body mass index; CED, Cohort entry date; COPD, chronic obstructive pulmonary disease; DVT, deep vein thrombosis, GP, general practitioner; PE, pulmonary embolism, RAS, renin-angiotensin- system; SD, standard deviation; Std Diff, absolute standardised difference;
Statistical analyses
To control for confounding, we applied inverse probability weighting by fine strata of the propensity score (PS) [18]. We calculated a PS based on all 32 baseline covariables (Table 1). We formed 50 PS strata based on the PS distribution in the cohort of diagnosed depression and weighted the cohort of patients with no diagnosed depression proportionally to the distribution of the exposed cohort in the PS stratum in which it fell. We trimmed the non-overlapping portions of the PS distributions to exclude non-comparable patients. The absolute standardized differences before and after weighting were used to assess balance of covariables. To obtain the measure of overall balance of covariables, we quantified the crude and weighted C-statistic (ideal balance c = 0.5) [19]. We assessed the distribution of baseline covariables for patients with diagnosed depression and patients with no diagnosed depression before and after PS weighting. We calculated crude and PS-weighted incidence rates (IRs) with 95% confidence intervals (CIs) for the two different cohorts separately. In addition, we quantified crude and PS-weighted IR ratios (IRRs) with 95% CIs for the four different types of infections by comparing patients with diagnosed depression to patients with not diagnosed depression. We calculated IRRs using negative binomial regression [20]. Whenever the negative binomial model did not converge, we used Poisson regression corrected for over dispersion [21].
Subgroup and sensitivity analyses
We performed analyses in subgroups of sex and age (18–24, 25–34, 35–44, 45–54, 55–64, and 65 + years).
To evaluate potential confounding by health care utilization (surveillance bias), we conducted subgroup analyses stratified by number of GP contacts within one year before cohort entry date (0–2, 3–5, 6–9,10–19, and 20 + GP visits). In an additional analysis, we removed the requirement for patients without a depression diagnosis to have GP contact within 14 days before cohort entry date. We performed this additional analysis to evaluate the possible influence of surveillance bias.
Depression may be a secondary consequence of chronic illness [22, 23]. Many chronic illnesses are also known risk factors for acute infections [24–26]. To evaluate potential confounding by underlying chronic illnesses, we performed a sensitivity analysis in which we excluded all patients with a record for any of the 14 baseline comorbidities within one year prior to cohort entry date, or with one or more prescriptions for any of 12 comedications within 180 days prior to the cohort entry date.
To validate the robustness of our outcomes, we performed two additional analyses with varying minimum time periods between separate episodes of infections of 7 and 28 days.
To assess whether a history of cancer and/or immunosuppression more than 2 years prior to cohort entry date is associated with susceptibility to acute infections, we performed a sensitivity analysis in which we excluded all patients who had a history of cancer and/or were prescribed an immunosuppressant at any time before cohort entry date (patients with a record for cancer or immunosuppression < = 2 years prior to the cohort entry date were already excluded).
To evaluate the potential role of severity of depression, we conducted an additional analysis in which we sub-divided the exposed group into two subgroups, treated or untreated depression, and compared them each to patients with no diagnosed depression. We censored follow-up for treated depression 180 days after the last antidepressant prescription. We stopped follow-up 180 days after the last depression diagnosis in patients with untreated depression.
Among patients with HES linkage, we calculated the proportion of patients with hospitalized infections in the exposed and non-exposed comparison groups separately. Additionally, we compared baseline characteristics between patients with or with no diagnosed depression who had a hospital-diagnosed infection.
All analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC).
Results
After weighting, we included 285 922 patients with first ever diagnosed depression who met our study criteria and matched 285 921 patients with no diagnosed depression. The mean age at cohort entry was 38.9 years (SD; 17.5) for patients with diagnosed depression and 39.3 years (SD; 17.5) in the comparison group, and 62.1% were female. In total, 66.6% of patients with diagnosed depression received a prescription for an antidepressant on the same day as the depression diagnosis (Flow Chart in Fig. 1). All baseline covariables were well balanced after PS-weighting (weighted c-statistic = 0.50, Table 1).
Fig. 1.
Flowchart of study population
The crude and PS-weighted IRs and IRRs for the primary and secondary outcomes are shown in Table 2. In the results, we focused on the weighted outcomes. Crude results are displayed in Tables 2 and 3, and S1; overall they did not differ materially from PS-weighted results.
Table 2.
Incidence Rates (95% CI) and Incidence Rates Ratios (95% CI) of diagnosed acute respiratory, genitourinary, gastrointestinal infections and septicaemia among patients with and without first ever diagnosed depression
| No. of Events | Total person-years of follow-up | Incidence rate per 1000 person-years (95% CI) | IRR (95% CI) | |
|---|---|---|---|---|
| Any acute infection | ||||
| No diagnosed depression | ||||
| Crude | 42 349 | 486 551.1 | 87.0 (86.2–87.9) | ref |
| PS-weighted | 40 840 | 488 121.5 | 83.7 (82.9–84.5) | |
| Diagnosed depression | ||||
| Crude | 29 949 | 292 940.0 | 102.2 (101.1–103.4) | 1.17 (1.16–1.19) |
| PS-weighted | 29 949 | 292 939.0 | 102.2 (101.1–103.4) | 1.19 (1.17–1.21) |
| Acute respiratory infections | ||||
| No diagnosed depression | ||||
| Crude | 26 389 | 486 551.1 | 54.2 (53.6–54.9) | ref |
| PS-weighted | 25 468 | 488 121.5 | 52.2 (51.5–52.8) | |
| Diagnosed depression | ||||
| Crude | 18 175 | 292 940.0 | 62.0 (61.1–63.0) | 1.14 (1.12–1.17) |
| PS-weighted | 18 175 | 292 939.0 | 62.0 (61.1–63.0) | 1.15 (1.13–1.18) |
| Acute genitourinary infections | ||||
| No diagnosed depression | ||||
| Crude | 14 976 | 486 551.1 | 30.8 (30.3–31.3) | ref |
| PS-weighted | 14 439 | 488 121.5 | 29.6 (29.1–30.1) | |
| Diagnosed depression | ||||
| Crude | 11 011 | 292 940.0 | 37.6 (36.9–38.3) | 1.22 (1.18–1.26) |
| PS-weighted | 11 011 | 292 939.0 | 35.6 (35.0–36.3) | 1.24 (1.20–1.28) |
| Acute gastrointestinal infections | ||||
| No diagnosed depression | ||||
| Crude | 863 | 486 551.1 | 1.8 (1.7–1.9) | ref |
| PS-weighted | 823 | 488 121.5 | 1.7 (1.6–1.8) | |
| Diagnosed depression | ||||
| Crude | 681 | 292 940.0 | 2.3 (2.2–2.5) | 1.31 (1.18–1.45) |
| PS-weighted | 681 | 292 939.0 | 2.2 (2.0–2.4) | 1.31 (1.19–1.45) |
| Acute septicaemia | ||||
| No diagnosed depression | ||||
| Crude | 239 | 486 551.1 | 0.5 (0.4–0.6) | ref |
| PS-weighted | 224 | 488 121.5 | 0.5 (0.4–0.5) | |
| Diagnosed depression | ||||
| Crude | 175 | 292 940.0 | 0.6 (0.5–0.7) | 1.22 (0.99–1.49) |
| PS-weighted | 175 | 292 939.0 | 0.6 (0.5–0.7) | 1.23 (1.00–1.50) |
Abbreviations: CI, confidence interval; IR, incidence rate; IRR incidence rate ratio; PS, propensity score;
Table 3.
IRR for acute infections comparing patients with first ever diagnosed depression to patients with no diagnosed depression restricted to patients without any recorded study comorbidities
| Variable | IRR (95% CI) | |
|---|---|---|
| Crude | PS-weighted | |
| Without recorded comorbidities or comedications | ||
| Any acute infection | 1.06 (1.03–1.09) | 1.09 (1.07–1.12) |
| Acute respiratory infections | 1.07 (1.04–1.10) | 1.10 (1.06–1.13) |
| Acute genitourinary infections | 1.04 (0.99–1.08) | 1.07 (1.03–1.12) |
| Acute gastrointestinal infections | 1.20 (1.03–1.40) | 1.20 (1.04–1.40) |
| Acute septicaemia | 1.15 (0.79–1.67) | 1.15 (0.79–1.67) |
After PS-weighting, 29 949 acute infections (primary outcome) were recorded during 292 939 person-years (py) of follow-up for patients with first ever diagnosed depression, resulting in an IR of 97.3 acute infections/1000 py. In the comparison cohort of patients with no diagnosed depression, 40 840 acute infections were recorded during 488 121.5 py, yielding an IR of 83.7 /1000 py. The weighted IRR for all acute infections comparing patients with first ever diagnosed depression to patients with no diagnosed depression was 1.19 (95% CI, 1.17–1.21) (Table 2). Exclusion of patients with a history of cancer (IRR 1.19, 95% CI, 1.17–1.21), or lengthening (IRR 1.19, 95% CI, 1.19–1.21) or shortening (IRR 1.19, 95% CI, 1.17–1.21) the minimum time period between separate outcome events of infection did not meaningfully change our results. (Table S1)
Detailed results of the subgroup and sensitivity analyses can be found in Tables 3 and 4; Fig. 2, and in Supplementary Tables S1-S6.
Table 4.
PS-weighted IRR for acute infections in patients with untreated or treated diagnosed depression compared with patients with no diagnosed depression
| Untreated diagnosed depression vs. no diagnosed depression | Treated diagnosed depression vs. no diagnosed depression | |
|---|---|---|
| PS Weighted IRR (95% CI) | ||
| Any acute infection | 1.03 (1.00–1.06) | 1.25 (1.22–1.27) |
| Acute respiratory infections | 1.03 (0.99–1.06) | 1.21 (1.18–1.24) |
| Acute genitourinary infections | 0.75 (0.73–0.77)* | 1.30 (1.25–1.34) |
| Acute gastrointestinal infections | 1.14 (0.96–1.36) | 1.43 (1.27–1.60) |
| Acute septicaemia | 1.27 (0.91–1.76) | 1.14 (0.89–1.44) |
Abbreviations: CI, confidence interval; IR, incidence rate; IRR, incidence rate ratios; PS, propensity score; *, POISSON
Fig. 2.
Result of subgroup analyses for the risk of any acute infection comparing patients with first ever diagnosed depression to patients with no diagnosed depression after PS-weighting
Incidence rates for the composite outcome of acute infection were highest (123.1/1000py) in patients under 25 years of age and lowest in patients aged 55–65 years. Despite the high absolute risk of infections in patients under 25 years of age, the weighted IRR was not materially increased (1.15, 95% CI, 1.12–1.19), whereas the weighted IRR for any acute infection was highest (1.50, 95% CI, 1.40–1.60) in the subgroup of patients aged 65 years and older. (Fig. 2)
The IR of new diagnosed infection was three times higher in women with first ever diagnosed depression (IR 127.6/1000py) compared to men (IR 46.9/1000py). (Fig. 2) This increase was mainly accounted for by genitourinary infections (IR 50.2/1000py in women and 11.2/1000py in men). However, IRRs for the composite outcome were not meaningfully different between women (IRR 1.18, 95% CI, 1.16–1.21) and men (IRR 1.10, 95% Cl, 1.06–1.15). (Fig. 2)
The IRRs of acute infections were similar across all strata of number of GP visits (Fig. 2).
In the additional analysis, in which the requirement to have a recorded GP visit at least once in the 14 days before cohort entry date was removed for patients without a depression diagnosis, the IRR for the primary outcome was significantly increased in patients with diagnosed depression compared to patients with no diagnosed depression (crude IRR, 1.60; 95% CI, 1.59–1.61, PS-weighted IRR 1.54; 95% CI, 1.52–1.55).
When we restricted the analysis to patients with no recorded diagnoses of any of the 14 baseline comorbidities at any time prior to the cohort entry date, and no prescriptions for any of 12 comedications within 180 days before cohort entry date, the IRR for any acute infection in patients with first ever diagnosed depression compared to no diagnosed depression was almost null: weighted IRR 1.09 (95% CI, 1.07–1.12) (Table 3). The same pattern was seen for all four primary outcomes when evaluated separately (Table 3).
The PS-weighted IRR for any acute infection in treated depression compared to no diagnosed depression was 1.25 (95% CI, 1.22–1.27). When we compared patients with untreated diagnosed depression to patients with no diagnosed depression, the PS-weighted IRR was 1.03 (95% CI, 1.00-1.06) for the composite outcome of any infection (Table 4).
The number of patients in our study population with a hospital-diagnosed infection was too small to yield interpretable results.
Discussion
In this large retrospective cohort study, there was no meaningful difference in risk of infection between patients with first ever diagnosed depression compared to patients with no diagnosed depression after controlling for confounding and surveillance bias. Excluding patients with a history of one or more recorded comorbidity or comedication before cohort entry further attenuated the initially observed minimal excess risk of acute infection (IRR 1.09). Thus, our results suggest that the risk of infection in association with diagnosed depression is lower than previously reported. The difference in results may be explained by control for underlying comorbidities and surveillance bias among patients with secondary depression.
The IRR of 1.19 in this study is lower than the IRR of 1.61 (95% CI, 1.49–1.74) reported in a previous national Danish registry-based cohort study [11]. The Danish study evaluated infections diagnosed in inpatient and outpatient settings. Results were adjusted for age and sex, but not for health care utilization or any other potential confounders. As not all patients seek medical attention for infection, patients who are in closer contact with the health care system (e.g. patients with diagnosed depression) are more likely to have an infection diagnosed than patients who do not see their GP regularly, potentially introducing surveillance bias. We controlled for the number of GP contacts within a year before cohort entry and required exposed and comparison patients to have at least one recorded GP visit within 14 days prior to cohort entry. In the analysis in which we dropped the latter requirement for non-exposed patients, the adjusted IRR increased to 1.54 for acute infection in patients with diagnosed depression compared to no diagnosed depression. This risk estimate is similar to the result in the Danish study, suggesting that their results may be explained by differential health care utilization. Prior studies have shown, that the lack of control for health care utilization between comparison groups may introduce bias [27].
This hypothesis is further supported by the fact that relative risks moved closer to the null after restricting to patients with no recorded study comorbidities or comedications. After this restriction we observed an IRR for any infection of interest of 1.09 (95% CI 1.07–1.12) and for genitourinary infections of 1.07 (95% CI, 1.03–1.12. (Table 3). Unlike common acute respiratory and gastrointestinal infections, urinary tract infections (UTIs) commonly require antibiotic treatment and are thus likely to be more completely reported to the GP. Thus, surveillance bias can be expected to play less of a role when analyzing this outcome compared to the other infections in this study.
A previous retrospective study analyzed survey data from the 2004 American College Health Association-National College Health Assessment (ACHA-NCHA) and reported a higher risk of any infection, with ORs ranging from 1.10 for ear infection to 1.30 for bronchitis among college students with self-reported depression compared to college students without self-reported depression. However, only a few potential confounding factors were controlled for in this survey study. Furthermore, college students who all lived on campus were substantially different than the adult primary-care population in this study, limiting comparability [10]. In addition, survey studies may be affected by selection (volunteer) bias compared to population bases studies such as this one.
Patients with first ever diagnosed depression may differ from patients with no diagnosed depression in many ways. Therefore, inferring a causal association between diagnosed depression and infection must be done with caution. To explore the role of potential confounding factors, we conducted several subgroup and sensitivity analyses. To evaluate the role of underlying comorbidities, which may have led to secondary depression, we excluded all patients with a recorded diagnosis for any of the 14 evaluated pre-existing comorbidities or for a co-medication within 180 days prior to the cohort entry date. Among the 55% of remaining patients with potentially primary diagnosed depression, the excess relative risk of developing an acute infection dropped from 19 to 9% compared to no diagnosed depression, suggesting that the severity of underlying comorbidities may have introduced residual confounding and that the risk of acute infection may be at least partially mediated by the underlying health status of patient with diagnosed depression.
To evaluate the role of the severity of diagnosed depression, we categorized patients with first ever diagnosed depression into those with or without a recorded prescription for an antidepressant on the day of first diagnosis. We assumed that patients receiving an antidepressant suffer from more severe depression. When comparing patients without an antidepressant, we no longer observed an increased risk of acute infection after PS-weighting (IRR, 1.03; 95% CI, 1.00-1.06). In contrast, the risk of acute infection was minimally increased for patients with diagnosed depression and pharmacological antidepressant therapy (PS-weighted IRR, 1.25; 95% CI, 1.22–1.27). After excluding patients with underlying comorbidities, the IRR decreased to 1.09 (95% CI, 1.107–1.12), suggesting that the severity of underlying pre-existing diseases may play a role in this minimal increase in risk.
We found a threefold increased absolute risk of acute infection for women compared to men. This difference is accounted for primarily by genitourinary infections. Genitourinary infections are much more common in women, and since most UTIs require antibiotic treatment, they are often recorded by GPs. However, the relative risk of acute infections overall was not meaningfully different between men and women in our study population.
Patients with diagnosed depression aged 65 years or older had an IRR of 1.5 for acute infection compared to patients with no diagnosed depression in the same age group. However, 88% of patients with diagnosed depression in this age group had at least one recorded pre-existing comorbidity, which may have confounded the result. When we excluded patients 65 or older with pre-existing comorbidities, the IRR of any acute infection decreased to 1.30 (95% CI of 1.02–1.67).
Strengths and limitations
Our cohort study was based on a large and representative study population and used rigorous methodology including several sensitivity and subgroup-analyses to comprehensively assess the role of confounding and bias. PS-fine stratification allowed us to adjust for recorded differences between cohorts.
However, some limitations need to be considered. First, GP-recorded IRs especially of acute respiratory and gastrointestinal infections do not reflect the true IR of such infections in the general UK population. Mild infections are generally not recorded if patients do not seek medical advice, which is common for infections such as acute respiratory and gastrointestinal infections, which often do not require GP-prescribed pharmacological treatment. Genitourinary infections and septicaemia are likely more completely entered in the GP record because they are more severe and typically require antibiotic therapy [28–30]. Severe infections (i.e. septicaemia and hospital recorded infections) were sparse and sample size was too small to be informative. Thus, our results address only diagnosed depression and moderate infections, which led to a contact with the GP but were not severe. Secondly, this study did not encompass all types of infections. We focused on four types of relatively frequently diagnosed infections that and are typically encountered in GP practices. We were also not able to differentiate between viral and bacterial infections, as this information is not routinely recorded in CPRD GOLD data. Thirdly, it is likely that some patients who have depression were undiagnosed. We can therefore not rule out some exposure misclassification where patients with undiagnosed depression are included in the control group. If the presence of depression does increase the risk of infection, this would have biased our results towards the null [31]. Moreover, receiving a diagnosis of depression per se may be associated with other factors that protect from infection when compared to patients with undiagnosed depression, which we could not control for. Fourth, it is possible that the cause for GP consultation in the unexposed cohort may have been associated with an increased risk of infection (e.g. diabetes, COPD) in a proportion of patients with no diagnosed depression. However, this likely did not meaningfully confound our results given that IRRs were closer to the null after excluding all patients with recorded comorbidities. We required comparison patients to have at least 1 GP visit within 2 weeks before CED, because prior studies have shown that omission to control for health care utilization between comparison groups introduces surveillance bias [27]. However, residual confounding cannot be completely ruled out.
Conclusions
This study suggests that patients with diagnosed depression are not at a meaningfully increased risk for the here examined acute infections when compared to patients with no diagnosed depression after adjusting for surveillance bias and comorbidities. Lack of control for health care seeking behaviour may explain the increased risks of infection in association with diagnosed depression seen in previous studies.
Electronic Supplementary Material
Below is the link to the electronic supplementary material.
Acknowledgements
We would like to thank everyone who helped with this study. Pascal Egger for his technical support and data processing, and Stephan Rueegg for his clinical insights.
Abbreviations
- CED
Cohort entry date
- CPRD
Clinical Practice Research Datalink
- HIV
Human immunodeficiency virus
- PS
Propensity score
Author contributions
All authors participated in the design of the study. In particular, Undine Lang provided the main contribution to the research question and Noah Aebi, Julia Spoendlin, Christoph Meier, and Susan Jick drafted the protocol. Noah Aebi prepared the lists of exposures and reading codes for the outcomes of interest and conducted the literature search. Julia Spoendlin was instrumental in the selection and application of the study design. Noah Aebi performed the statistical analysis, designed the tables, figures, and appendices, and prepared the first draft of the manuscript, with assistance from Julia Spoendlin. All authors critically reviewed and approved the final version of the manuscript. The corresponding author certifies that all listed authors meet the authorship criteria and that no other authors who meet the criteria have been omitted.
Funding
Open access funding provided by University of Basel
Data availability
No datasets were generated or analysed during the current study.
Declarations
Ethics approval and consent to participate
This study was approved by the Independent Scientific Advisory Committee for Medicines and Healthcare Products Regulatory Agency database research (protocol no: 21_000398) and has been made available to the journal editors.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Bromet E, Andrade LH, Hwang I, Sampson NA, Alonso J, de Girolamo G et al (2011) Cross-national epidemiology of DSM-IV major depressive episode.(Research article)(Survey). BMC Med [DOI] [PMC free article] [PubMed]
- 2.Malhi GS, Mann JJ, Depression (2018) Lancet [DOI] [PubMed]
- 3.Lopez AD, Mathers CD (2006) Measuring the global burden of disease and epidemiological transitions: 2002–2030. Ann Trop Med Parasitol [DOI] [PubMed]
- 4.Vos T, Allen C, Arora M, Barber RM, Brown A, Carter A et al (2016) Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet [DOI] [PMC free article] [PubMed]
- 5.Leonard BE (2010) The concept of depression as a dysfunction of the immune system. Curr Immunol Rev 6:205–212 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Leonard B (2000) Stress, depression and the activation of the immune system. world J Biol psychiatry Off J World Fed Soc Biol Psychiatry Engl 1:17–25 [DOI] [PubMed] [Google Scholar]
- 7.Dantzer R, O’Connor JC, Freund GG, Johnson RW, Kelley KW (2008) From inflammation to sickness and depression: when the immune system subjugates the brain. Nat Rev Neurosci 9:46–56 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Leach LS, Christensen H, Mackinnon AJ, Windsor TD, Butterworth P (2008) Gender differences in depression and anxiety across the adult lifespan: the role of psychosocial mediators. Soc Psychiatry Psychiatr Epidemiol Ger 43:983–998 [DOI] [PubMed] [Google Scholar]
- 9.Baddeley JL, Pennebaker JW, Beevers CG (2012) Everyday Social Behavior During a Major Depressive Episode. Soc Psychol Personal Sci. SAGE Publications Inc; ;4:445–52. 10.1177/1948550612461654
- 10.Adams TB, Wharton CM, Quilter L, Hirsch T (2008) The association between mental health and acute infectious illness among a national sample of 18- to 24-year-old college students. J Am Coll Heal [DOI] [PubMed]
- 11.Andersson NW, Goodwin RD, Okkels N, Gustafsson LN, Taha F, Cole SW et al (2016) Depression and the risk of severe infections: Prospective analyses on a nationwide representative sample. Int J Epidemiol [DOI] [PubMed]
- 12.Rantala M, van de Laar MJW (2008) Surveillance and epidemiology of hepatitis B and C in Europe - a review. Euro Surveill Bull Eur sur les Mal Transm = Eur Commun Dis Bull Swed; 13 [DOI] [PubMed]
- 13.Hahné S, Ramsay M, Balogun K, Edmunds WJ, Mortimer P (2004) Incidence and routes of transmission of hepatitis B virus in England and Wales, 1995–2000: implications for immunisation policy. J Clin Virol Off Publ Pan Am Soc Clin Virol Neth 29:211–220 [DOI] [PubMed] [Google Scholar]
- 14.Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, van Staa T et al (2015) Data Resource Profile: Clinical Practice Research Datalink (CPRD). Int J Epidemiol 44:827–836 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Herrett E, Thomas SL, Schoonen WM, Smeeth L, Hall AJ (2010) Validation and validity of diagnoses in the General Practice Research Database: a systematic review. Br J Clin Pharmacol 69:4–14 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Jick SS, Kaye JA, Vasilakis-Scaramozza C, Garcia Rodríguez LA, Ruigómez A, Meier CR et al (2003) Validity of the general practice research database. Pharmacotherapy United States 23:686–689 [DOI] [PubMed] [Google Scholar]
- 17.Thomas KH, Davies N, Metcalfe C, Windmeijer F, Martin RM, Gunnell D (2013) Validation of suicide and self-harm records in the Clinical Practice Research Datalink. Br J Clin Pharmacol 76:145–157 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Desai RJ, Franklin JM (2019) Alternative approaches for confounding adjustment in observational studies using weighting based on the propensity score: A primer for practitioners. BMJ [DOI] [PubMed]
- 19.Franklin JM, Rassen JA, Ackermann D, Bartels DB, Schneeweiss S (2014) Metrics for covariate balance in cohort studies of causal effects. Stat Med Engl 33:1685–1699 [DOI] [PubMed] [Google Scholar]
- 20.Gardner W, Mulvey EP, Shaw EC (1995) Regression analyses of counts and rates: Poisson, overdispersed Poisson, and negative binomial models. Psychol Bull United States 118:392–404 [DOI] [PubMed] [Google Scholar]
- 21.Kang EH, Jin Y, Desai RJ, Liu J, Sparks JA, Kim SC (2020) Risk of exacerbation of pulmonary comorbidities in patients with rheumatoid arthritis after initiation of abatacept versus TNF inhibitors: A cohort study. Semin Arthritis Rheum United States 50:401–408 [DOI] [PubMed] [Google Scholar]
- 22.DeJean D, Giacomini M, Vanstone M, Brundisini F (2013) Patient experiences of depression and anxiety with chronic disease: a systematic review and qualitative meta-synthesis. Ont Health Technol Assess Ser 13:1–33 [PMC free article] [PubMed] [Google Scholar]
- 23.Read JR, Sharpe L, Modini M, Dear BF (2017) Multimorbidity and depression: A systematic review and meta-analysis. J Affect Disord Neth 221:36–46 [DOI] [PubMed] [Google Scholar]
- 24.Frydrych LM, Bian G, O’Lone DE, Ward PA, Delano MJ (2018) Obesity and type 2 diabetes mellitus drive immune dysfunction, infection development, and sepsis mortality. J Leukoc Biol United States 104:525–534 [DOI] [PubMed] [Google Scholar]
- 25.Kim EJ, Ha KH, Kim DJ, Choi YH (2019) Diabetes and the Risk of Infection: A National Cohort Study. Diabetes Metab J 43:804–814 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Fünfstück R, Nicolle LE, Hanefeld M, Naber KG (2012) Urinary tract infection in patients with diabetes mellitus. Clin Nephrol Ger 77:40–48 [DOI] [PubMed] [Google Scholar]
- 27.Schneeweiss S, Patrick AR, Stürmer T, Brookhart MA, Avorn J, Maclure M et al (2007) Increasing levels of restriction in pharmacoepidemiologic database studies of elderly and comparison with randomized trial results. Med Care United States 45:S131–S142 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Monto AS (2002) Epidemiology of viral respiratory infections. Am J Med. 112:4–12. https://www.sciencedirect.com/science/article/pii/S0002934301010580 [DOI] [PubMed]
- 29.Foxman B (2010) The epidemiology of urinary tract infection. Nat Rev Urol [DOI] [PubMed]
- 30.Foxman B (2002) Epidemiology of urinary tract infections: incidence, morbidity, and economic costs. Am J Med. 113:5–13. https://www.sciencedirect.com/science/article/pii/S0002934302010549 [DOI] [PubMed]
- 31.Kessler D, Lloyd K, Lewis G, Gray DP, Heath I (1999) Cross sectional study of symptom attribution and recognition of depression and anxiety in primary care Commentary: There must be limits to the medicalisation of human distress. BMJ [DOI] [PMC free article] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
No datasets were generated or analysed during the current study.