Loop diuretic utilisation with or without heart failure: impact on prognosis

Many patients are prescribed loop diuretics without a diagnostic record of heart failure. Little is 3 known about their characteristics and prognosis.


INTRODUCTION
Heart failure (HF) is characterised by water and salt retention leading to congestion of the systemic and pulmonary circulation and eventually to symptoms, such as exertional breathlessness, and signs, such as peripheral oedema (1,2).Guidelines on HF strongly recommend loop diuretics for managing symptoms and signs of congestion (1,3,4), but loop diuretics can also be used for resistant hypertension or managing congestion due to end-stage kidney disease (1).For patients with HF, the development of congestion and the need for loop diuretics to manage it are associated with an adverse prognosis (2).However, many patients are prescribed loop diuretics but are not investigated for the underlying cause; if they relieve symptoms and signs, they may mask a diagnosis of HF (1).
To find out how many patients are prescribed loop diuretics, how often this is associated with a diagnosis of HF and whether prescribing loop diuretics is associated with an adverse prognosis, electronic health records were obtained for patients in the Greater Glasgow & Clyde region if they either had a diagnosis of coronary or peripheral arterial disease or HF or were prescribed loop diuretics or medicines commonly prescribed for hypertension or ventricular dysfunction.

Data sources
In Scotland, every resident has free access to primary and secondary healthcare, and free prescriptions through the National Health Service (NHS).Residents receive a unique identification number, and all healthcare contacts and deaths are linked through this.For people served by the NHS Greater Glasgow & Clyde Health Board, person-level pseudonymised administrative data were obtained, including demographic data, primary care diagnostic information; hospital admissions and associated diagnostic and procedural codes; communitybased prescriptions (5); and death records (Supplementary Table 3 for data sources).Any haematology and biochemistry test results from primary and secondary care and reports of electrocardiograms (ECGs) and echocardiograms were also obtained.Data extraction and Patients were eligible for inclusion if they were alive and aged ≥18 years on 1 st January 2012, had records available for ≥12 months and had a record of a diagnosis of coronary artery disease (CAD), peripheral arterial disease or HF between 31 st December 2009 and 31 st    December 2011, or if, for any reason including hypertension, they were dispensed, angiotensinconverting enzyme inhibitors (ACEi), angiotensin II receptor blockers (ARB), beta-blockers, mineralocorticoid receptor antagonists (MRA), or loop diuretics.Diagnoses were identified using either the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) in any position or primary care Read Codes, both with records going back to the year 2000 (Supplementary Table 3).For patients who were included, primary and secondary care diagnostic records were made available from the year 2000 onwards, although prescribing data were not available before 31 st December 2009.Medicines dispensed from community pharmacies were identified using British National Formulary codes.

Patient identification and categorisation
Treatment with loop diuretics (Supplementary Table 7 for list) was defined as the first occasion that loop diuretics were both prescribed and dispensed in two consecutive quarters or if death occurred within 90 days of first prescription.HF was defined as a relevant diagnostic record in either primary care (Read codes) or a hospitalisation discharge code in any of six possible diagnostic positions (ICD-10 codes) from the year 2000 onwards.Code lists were developed using previously published research (6) and through searching for the term 'heart failure' in the NHS coding dictionary (Supplementary Table 4).Evidence of structural heart disease was not required but relevant investigations are reported where available.Following previous conventions (6), some patients (n = 627) were excluded because of uncertainty about HF diagnosis date (Supplementary Table 5; Supplementary Figure 1).A first diagnosis of HF during a fatal hospitalisation (1,093 deaths) was not counted as incident HF because of diagnostic uncertainty and lack of information on in-hospital loop diuretic use.
Patients were classified on 31 st December 2011 into one of four mutually exclusive groups based on a prior record of HF at any time in the prior decade and recurrent dispensing of loop diuretics in the prior 6 months, as: i) neither HF nor on loop diuretic, ii) loop diuretic only, iii) HF only, or iv) HF on loop diuretics.Because the average age of patients in the "neither" group was considerably younger than the other three groups, it was further divided into those aged 18-59 years or ≥60 years.

Study outcomes
Patients were followed from 1 st January 2012 until the 31 st December 2016.Follow-up was right censored on the date of the last available health record, including blood tests or other investigations or dispensing, to avoid uncertainty about survival status if patients emigrated out of the region as this would otherwise introduce an 'immortal' bias.The main outcome of interest was 5-year all-cause mortality.The primary covariate of interest was group membership defined by loop diuretic dispensing and HF diagnosis.The 5-year cumulative HF incidence, loop diuretic initiation, cause-specific hospitalisation and mortality were also investigated.Causes of hospitalisation were defined by the primary discharge code and mapped onto twelve ICD-10 disease categories.Cause of death was defined from patient records and classified as cardiovascular, neoplastic, infection, other, and unknown (Supplementary Table 9).Co-morbidities in the decade prior to 1 st January 2012 were identified from primary and secondary care records, and included hypertension, diabetes mellitus, thyroid disease, atrial fibrillation, or flutter (AF), CAD (including myocardial infarction), valve disease, peripheral arterial disease, stroke, chronic obstructive pulmonary disease (COPD), cancer, liver disease, and dementia.Diagnostic code lists were adapted from CALIBER phenotypes (8) and previously published research (9) (Supplementary Table 6).Patients without a diagnostic record for a specific disease were assumed to be free from that condition.
In addition to ACEi, ARB, beta-blockers, MRA and loop diuretics, information on calcium channel blockers, digoxin, thiazides and thiazide-related diuretics, aspirin, non-steroidal anti-inflammatory drugs, lipid regulators, bronchodilators, thyroid medications, and hypoglycaemic agents including insulin were reported.Patients were considered to be on these medications if dispensed in the 180 days before 1 st January 2012.Medicines were identified by British National Formulary codes (Supplementary Table 7).
The most recent haemoglobin and estimated glomerular filtration rate (eGFR) measured between 2010-2012 were reported.Anaemia was defined as <12.0 g/dL for women and <13.0 g/dL for men (10), and eGFR was calculated using the Chronic Kidney Disease-Epidemiology Collaboration equation (11) without adjusting for ethnicity (12).
Measurements of left ventricular ejection fraction (LVEF) and left atrial diameter were obtained from routinely available echocardiograms (if available) and heart rhythm and QRS duration from ECGs.Results closest to a diagnosis of HF or initiation of loop diuretics were chosen if there were multiple tests.LVEF was considered reduced if <50% and left atrium dilated if >4.0 cm for men and >3.8 cm for women (13).

Statistical analyses
Patient characteristics are presented as numbers and percentages for categorical variables and median with interquartile range (IQR) for continuous variables.Numbers and percentages of complete records are displayed for variables with missing data.Percentages of categorical variables refer to complete cases.
Prevalence was estimated within mid-year 2012 regional population estimates (14), using sex-stratified, 5-year age bands as the denominator from ages 29 to >90 years.
Admission rates were calculated as the number of admissions per patient-year at risk where the patient was at risk of being admitted (alive, not in hospital, and not lost to followup).Allowance was made for the competing risk of death when estimating cumulative incidences of initiating loop diuretics and incident HF.
A Cox proportional hazards regression model was used to assess between-group differences in all-cause mortality with a robust sandwich-type estimator due to the potential lack of statistical independence between chronic comorbidities (15).Due to the large sample size and high statistical power to detect small departures from proportional hazards, the proportional effects were visually checked using log-log plots.The model was adjusted for age, as a continuous, linear value, kidney function using a penalised spline on eGFR with five knots (Handling missing eGFR data in supplementary material), sex, Scottish Index of Multiple Deprivation, and history of common, chronic conditions: hypertension, CAD, peripheral arterial disease, diabetes mellitus, valve disease, AF, stroke, cancer, and dementia were selected based on clinical expertise.Results are reported as hazard ratios (HR) with 95% confidence intervals (CI).The estimated cumulative incidence of death due to cardiovascular disease, infection, neoplasm, or other causes of death, and all-cause mortality after initiation of loop diuretics or incident HF were also calculated from sub-distribution estimates.Time-dependent covariates were used to assess the impact of disease progression on morbidity and mortality.The dates of HF diagnosis and the start of repeated loop diuretic dispensing were used to update loop diuretic/HF group.If both events occurred on the same day, patients were classified as being on the combination to avoid immortal time bias (timedependent covariate analysis in supplementary material).Crude 5-year morbidity and mortality rates were calculated using person-time at risk, where attributable time was allocated based on time-dependent group status.All-cause mortality was modelled using Cox proportional hazards and time-dependent covariates, using time-dependent comorbidity and loop diuretic/HF group values and baseline age, sex, and Scottish Index of Multiple Deprivation.

Study findings are reported as advised by REporting of studies Conducted using
Observational Routinely-collected health Data recommendations (16).Data were prepared using Microsoft SQL Server Management Studio (version 17.8.1),and statistical analyses were performed using R (version 4.0.5).Relevant packages are listed in the supplementary material.198,898 met the cohort inclusion criteria, which comprised more than half of all men aged >70 years and more than half of all women aged >75 years (Figure 1).The cohort included 161,935 (81%) patients who had no record of HF in the prior decade and had not been dispensed loop diuretics in the prior 6 months (of whom 89,699 were aged ≥60 years), 23,963 (12%) who had been dispensed loop diuretics but had no record of HF, 7,844 (4%) who had HF and were dispensed loop diuretics, and 5,156 (3%) who had a record of HF but were not dispensed loop diuretics.The estimated prevalence of HF for the population (including an estimated 0.3 million people aged <18 years) was 1.0% overall, or 1.3% of the adult population (Figure 1).
Of 77,029 patients who were never subsequently initiated on loop diuretics and did not develop HF, LVEF was <50% in 443 (6%) of 7,221 patients with measurements and the left atrium was dilated in 5,884 (46%) of 12,915 patients with measurements (Supplementary Figures 3 and 4).More than two-thirds of patients were on an ACEi or ARB, 64% on lipidlowering agents, 45% on beta-blockers, 36% on calcium channel blockers, 35% on thiazides, and 45% on aspirin (Table 2).Patients aged 18-59 years who neither had HF recorded nor dispensed loop diuretics had fewer co-morbidities than any of the other groups.

A C C E P T E D
People dispensed loop diuretics but without a record of HF had a median age of 75 (IQR: 65-83) years, were predominantly women (70%), and many (44%) were in the most socio-economically deprived quintile.Compared to those aged ≥60 years in the 'neither' group, they had a similar prevalence of hypertension, diabetes mellitus, CAD, and myocardial infarction, but more COPD (19%), AF (16%), and anaemia (36%).Although many patients had an eGFR <60 mL/min/1.73m 2 (34%), it was rarely <30 mL/min/1.73m 2 (6%).Of the 31,872 patients who received loop diuretics at baseline or follow-up but never received a diagnosis of HF, LVEF was <50% in 509 (9%) of 5,409 patients with measurements and the left atrium was dilated in 5,548 (57%) of 9,669 patients with measurements.Compared to patients in the 'neither' group, patients were less likely to be prescribed ACEi or ARB or thiazide diuretics but more likely to receive bronchodilators and oral anticoagulants.Few patients (15%) received three or more anti-hypertensive agents (excluding loop diuretics), a criterion commonly used to define resistant hypertension (17).
People with HF who were not dispensed loop diuretics were younger (median age 69 (IQR: 59-78) years) and predominantly men (68%).Compared to patients taking loop diuretics without a diagnosis of HF, they had a high prevalence of CAD (75%) and myocardial infarction (53%) and were less likely to have an eGFR <60 mL/min/1.73m 2 (20%).Of 5,608 patients who had a diagnosis of HF at baseline or follow-up but who never received loop diuretics, LVEF was <50% in 381 (31%) of 1,212 patients with measurements and the left atrium was dilated in 1,112 (55%) of 2,022 patients with measurements.Compared to patients taking loop diuretics without a diagnosis of HF, patients with HF only were more likely to receive an ACEi or ARB (77%), beta-blockers (69%), lipid regulators (78%) and aspirin (66%) and less likely to receive bronchodilators (16%).

A C C E P T E D
Patients with a diagnosis of HF who were also treated with loop diuretics had a median age of 77 (IQR: 68-83) years, similar to other patients treated with loop diuretics, and 50% were women.Compared to patients taking loop diuretics without a diagnosis of HF, they were more likely to have a history of hypertension (52%), diabetes (29%), CAD (67%), myocardial infarction (40%), valve disease (20%), AF (45%), COPD (28%), and anaemia (45%).Many patients had an eGFR <60 mL/min/1.73m 2 (48%), but it was rarely <30 mL/min/1.73m 2 (9%).Of the 14,938 patients with a diagnosis of HF at baseline or follow-up who also received loop diuretics, LVEF was <50% in 1,701 (40%) of 4,279 patients with measurements and the left atrium was dilated in 5,545 (74%) of 7,518 patients with measurements.Compared to patients taking loop diuretics without a record of HF, they were more likely to receive an ACEi or ARB (74%), beta-blockers (64%), digoxin (23%), and oral anticoagulants (29%), and less likely to receive calcium channel blockers (22%), but only a minority were on MRA (15%).
Initiation of loop diuretics and incident heart failure during follow-up.
Over the following 5 years, for patients in the 'neither' group aged ≥60 years, 9,706 (10.8%) started taking loop diuretics, of whom only 12.6% subsequently had HF recorded (Figure 2).A further 2,951 (3.3%) patients received a diagnosis of HF, of whom 1,728 (58.7%) were later initiated on loop diuretics.Overall, 4,191 (4.7%) were diagnosed with HF before or after initiation of loop diuretics.For patients initially in the 'neither' group and aged ≥60 years (all of whom had evidence of or were receiving treatment for cardiovascular disease), 14,062 (15.7%) died over the following 5 years without receiving a diagnosis of HF or being dispensed loop diuretics.Of those initiated on loop diuretics who did not receive a diagnosis of HF, 2,640 (27.2%) subsequently died, and of those initiated on loop diuretics who did receive a diagnosis of HF, 1,040 (35.0%) subsequently died (Figure 3 and Supplementary Figure 7).A similar pattern, although at much lower rates, was observed for patients with cardiovascular disease aged 18-59 years (Figure 2).Very few deaths (5.7%) occurred within 90 days of first loop diuretic prescription that might have prevented appropriate, timely investigation.Of 23,963 patients taking loop diuretics at baseline but without a diagnosis of HF, only 2,635 (11.0%) subsequently received a HF diagnosis of whom 1,375 (52.2%) died; 8,266 (34.5%) died without first getting a HF diagnosis (Figure 2).Of 5,156 patients with HF not receiving loop diuretics at baseline, only 1,119 (21.7%) were later initiated on loop diuretics of whom 325 (29.0%) died; 812 (15.7%) patients died without being dispensed loop diuretics.For those with HF taking loop diuretics at baseline, mortality was 52.3% (Figure 2).
In the overall cohort, 54.4% of deaths were preceded by initiation of loop diuretics or a diagnosis of HF (Supplementary Figure 16).The most common source of new HF diagnosis was hospitalisations.Loop diuretic initiation without a diagnosis of HF generally occurred in primary care (55.0% of initiations) with no secondary care contact (ward or clinic) in the prior 30 days.

Hospitalisations
Compared to patients aged ≥60 years who did not have HF and were not taking loop diuretics, rates of hospital admission per patient-year at risk were higher for those taking loop diuretics whether or not they had a diagnosis of HF (Figure 4).In the 5 years after 1 st January 2012, 39,341 (54.5%) patients in the 'neither' group aged 18-59 had at least one admission, with 131,664 admissions in all (0.4 per patient-year at risk) and 66,124 (73.7%) patients in the  4).However, only a minority of admissions for all groups of patients was attributed to cardiovascular disease and an even smaller proportion to HF. Timedependent analysis showed that rates of admissions increased based on updated HF and loop diuretic status (Supplementary Figures 12 and 13), especially for cardiovascular and infectionrelated admissions.

Mortality
For mortality, the proportional hazards assumption was not met for the first 14 days, but this applied to only 328 deaths (0.9% of all deaths).Accordingly, proportional hazards were considered a reasonable summary of between-group differences.Using the entire "neither" loop diuretics nor HF group as reference, the adjusted HR (which includes age) for 5-year allcause mortality for those dispensed loop diuretics without a diagnosis of HF was 1.8 (95% CI: 1.7-1.8),for those with HF who were not dispensed loop diuretics was 1.2 (95% CI: 1.1-1.3),and for HF treated with loop diuretics was 2.1 (95% CI: 2.0-2.2) (Figure 5).In the absence of loop diuretics, thiazide diuretics were not associated with adverse outcomes.Applying timedependent covariates increased the strength of the association between loop diuretics and mortality, especially for patients with HF (Supplementary Figure 14).
Using baseline loop diuretic/HF classification, all-cause and cardiovascular mortality at five years for patients who had neither HF nor were prescribed loop diuretics and were aged <60 years were, respectively, 3.3% and 0.8%, and for those ≥60 years 20.2% and 5.8%, for patients with a diagnosis of HF who were not taking loop diuretics it was 22.1% and 8.2%, for those dispensed loop diuretics without a HF diagnosis it was 40.2% and 12.1% and for those with HF taking loop diuretics 52.3% and 22.6% (Figure 2; and Supplementary Figures 8 and 9, and Table 12).There were similar rates of deaths due to neoplasms for each of these three groups (Supplementary Figures 8 and 15).Of 23,963 patients in the loop diuretic-only group, 6,506 (79.9% women) were included solely because they were dispensed loop diuretics without meeting other criteria.Of these, 643 had a history of CAD or peripheral arterial disease before 31st December 2009.Of the remaining 5,863, 2,251 (38.4%) died.This compares with an all-cause mortality of 40.2% (7,013 deaths) in the remaining 17,457 patients who were included because they had cardiovascular disease or were prescribed other medications.

DISCUSSION
This analysis suggests that for patients with a broad range of cardiovascular diseases, mortality is more closely associated with taking loop diuretics than with a diagnosis of HF, even after adjusting for age and other risk factors (Structured Graphical Abstract).The estimated prevalence of HF amongst adults in Glasgow of 1.3% is consistent with data from elsewhere in the United Kingdom (6) but many more patients (3.2%) were dispensed loop diuretics.Only one in four patients treated with loop diuretics had a diagnostic record of HF, and only 11.0% were subsequently diagnosed with HF over the following 5 years.The prognosis of patients treated with loop diuretics, even without a diagnosis of HF, was substantially worse than that of patients with HF who were not receiving loop diuretics.Indeed, patients with a diagnosis of HF who were not receiving loop diuretics had only a slightly worse prognosis than patients aged ≥60 years with a broad range of other cardiovascular problems.Patients with HF who were treated with loop diuretics had the worst prognosis.Hospitalisation rates were also higher for patients taking loop diuretics with or without a diagnosis of HF, although the primary reason was usually for conditions other than cardiovascular disease for all patient groups.In summary, patients dispensed loop diuretics constitute a much larger healthcare problem than patients with a diagnostic record of HF, which might be explained by substantial under-diagnosis or under-recording of HF (1).This has serious implications for HF epidemiology and health service capacity to manage it.Patients with cardiovascular disease treated with loop diuretics are at high risk even if they do have no record of HF, which should alert clinicians to consider the need for further investigation and treatment.
The observation that loop diuretics are associated with an adverse prognosis in the absence of HF is not unique to Glasgow.International trials of AF (18), and type 2 diabetes mellitus (19) show that patients treated with loop diuretics often do not carry a diagnosis of HF but have worse outcomes than those with a diagnostic label of HF who are not treated with loop diuretics; those with both HF and loop diuretics consistently have the worst outcome.A study of patients with AF from England found that those receiving loop diuretics but no record of HF had a prognosis similar to those with HF recorded (20).
Current criteria used to define HF are not robust, relying heavily on symptoms and signs such as breathlessness and ankle swelling, that lack specificity and may not be obvious until HF is severe, requiring the patients to be hospitalised (1,6,(21)(22)(23).In the current analysis, few patients initiated on loop diuretics without a diagnosis of HF had measured natriuretic peptides or an echocardiogram to support or exclude a diagnosis of HF.However, many patients had hypertension, diabetes mellitus, anaemia, AF, and impaired kidney function, common comorbidities that may cause or exacerbate HF.The diagnosis of HF in the presence of a preserved LVEF is especially difficult when other reasons for symptoms and signs exist, such as lung or kidney disease (24).Many patients treated with loop diuretics who were not given a diagnosis of HF were older women and had left atrial dilation, consistent with the demographic profile and diagnosis of HF with preserved ejection fraction.AF may also cause left atrial dilation, but if it leads to symptoms and signs of congestion requiring treatment with loop diuretics, it may also be considered to have caused HF.Such diagnostic uncertainties might explain why many patients are treated with loop diuretics but are not diagnosed with HF.About half of patients had been hospitalised in the year before initiation of loop diuretics, and many others attended hospital out-patients, usually for non-cardiovascular conditions.A few patients treated with loop diuretics had end-stage kidney disease, and some will have received loop diuretics for the treatment of resistant hypertension, but neither indication for loop diuretics appeared to account for their widespread use (3).On general medical and surgical wards, loop diuretics may have been initiated for symptoms and signs of HF, but appropriate investigations not done or not recorded in patients' medical records (25).
When loop diuretics are listed as hospital discharge medications, primary care physicians may automatically repeat the prescription, trusting that their hospital colleagues have investigated appropriately.Patients admitted with, for example, respiratory infections, may receive both antibiotics and loop diuretics due to diagnostic uncertainty.Even if antibiotics are responsible for the success of treatment, loop diuretics may be continued long-term because no one decided to stop them.
Current consensus guidelines state that symptoms and signs of congestion are essential diagnostic criteria for HF (21), and strongly recommend loop diuretics for their treatment (4).
In the current analysis, many patients with a diagnostic label of HF were not treated with loop diuretics.These patients had often had myocardial infarctions, had reduced LVEF and were prescribed renin-angiotensin system inhibitors and beta-blockers.The relatively favourable prognosis of patients with HF who were not treated with loop diuretics may reflect the effective deployment of guideline-recommended therapies that prevent or reverse the development of congestion, rendering treatment with loop diuretics unnecessary.However, the decision not to initiate loop diuretics suggests that these patients had few or no symptoms or signs of congestion, and may not have fulfilled current guideline criteria for a diagnosis of HF.
That there is an association between taking loop diuretics and an adverse prognosis in patients with HF is not surprising (26).Loop diuretics are used to treat symptoms and signs of congestion, which is associated with more severe cardiac and renal dysfunction and a higher mortality.For patients with severe congestion, loop diuretics are almost certainly life-saving in the short term, but the longer-term adverse consequences, including activation of neuroendocrine systems, electrolyte disturbances and increased calcium excretion, might increase morbidity and mortality (27)(28)(29).The overall prognosis of patients taking loop diuretics will reflect the average prognosis for each of the diverse reasons for their use.Patients who received loop diuretics for ankle swelling in the absence of serious underlying disease might have a good prognosis, and therefore, for other patients within this diverse group, it must be much worse.It is also possible that prognosis is driven primarily by problems other than HF, such as lung disease or cancer, for which loop diuretics might do more harm than good.
Further research is required to determine how often loop diuretic prescription is simply a bystanding marker of a poor outcome or associated with symptoms and signs due to cardiac dysfunction; in other words, HF.

Strengths and limitations
Administrative health records were accessed for a region encompassing 23% of the Scottish population ( 14) including all community-based prescriptions and blood tests from 2010 onwards and primary care diagnosis and reasons for hospital admissions from the year 2000 onwards.However, some important variables were unavailable, including height, weight, smoking habit, blood pressure, dose, and dose frequency.Unlike many other large administrative datasets (8, 30, 31), a large sample of echocardiographic and ECG results were available.However, many more tests may have been done at the "bedside" without results being entered into electronic records.In future, accessing patients' case notes may be possible to retrieve this information.Information on in-hospital prescribing of interventions such as intravenous loop diuretic use was not available.
The cohort was defined with a broad, though incomplete, set of cardiovascular diseases, missing those with conditions such as AF, valvular disease, or venous thromboembolism.However, any patient dispensed an ACEi, ARB, MRA, beta-blocker, or loop diuretic would still have been included.This might have influenced the risk of the reference group but will not have the relative amongst groups.
Hypertension was probably under-reported, given the high prescription rates for ACEi, ARB, calcium channel blockers, and thiazide diuretics.However, the prevalence of diabetes and percentage of patients on hypoglycaemic therapy and the prevalence of COPD and percentage on bronchodilators were similar, suggesting that these are useful pharmaco-epidemiological markers of disease (32,33), just as loop diuretics might be for symptoms and signs of HF.
Research using administrative health records is reliant on clinical coding.In an audit of Scottish hospitalisation records, the 'I50' code for HF was judged to be correct >90% of the time but missed a diagnosis of HF in 22% of cases (34).Death certificates are probably fairly accurate for classifying deaths as cardiovascular or due to cancer but may be less reliable for further specifying causes such as HF or sudden death.

A C C E P T E D
In conclusion, this analysis suggests that in patients with a broad range of cardiovascular diseases, mortality is more strongly associated with use of loop diuretics than with a diagnosis of HF, but amongst those treated with loop diuretics, a diagnosis of HF is associated with a worse prognosis.Either the diagnosis of HF is often missed, which may result in the withholding of evidence-based treatment of HF, or loop diuretic use is associated with other conditions with a prognosis similar to HF, or inappropriate loop diuretic use increases mortality; all might be true and contribute to poor outcomes.Nevertheless, treatment with loop diuretics provides a simple, reliably-collected, objective marker of patients at increased risk of hospitalisation and death.The proportion of patients on loop diuretics undergoing investigation for HF could be used to audit the quality of diagnostic care in clinical practice.When a clinician encounters a patient treated with loop diuretics, this should trigger a review of the patient's medical records to ensure that appropriate investigations have been done, such as measurement of natriuretic peptides or cardiac imaging, to exclude serious cardiac pathology.date to account for patients who moved out of the region).The total number of deaths (n) per group are reported above each column.Supplementary Figure 14 shows similar data with loop diuretic/HF group as a time-dependent covariate.
record linkage were performed by the West of Scotland Safe Haven service.Ethical approval was given by the West of Scotland Safe Haven's Local Privacy Advisory Committee, reference number GSH/18/CA/002.Study population.
for 1 st January 2012 included age, sex, ethnicity, Scottish Index of Multiple Deprivation 2012 quintile, an area-based measurement of socioeconomic deprivation where lower numbers indicate higher levels of deprivation (see supplementary materials) (7), co-morbidities, medication, blood tests, and, when available, the results of ECGs and echocardiograms.
982,385 adults in the NHS Greater Glasgow & Clyde region in 2012,

'
neither' group aged ≥60 years had 261,793 admissions (0.7 per patient-year at risk).Patients taking loop diuretics who did not have HF at baseline had 91,361 admissions over 5 years (1.0 per person-year at risk, with 19,884 [83.0%] having at least one admission).Patients with HF who were not dispensed loop diuretic had 18,317 admissions over 5 years (0.8 per person-year at risk, with 4,063 [78.8%] having at least one admission).Patients with HF taking loop diuretics had 34,120 admissions over 5 years (1.3 per person-year at risk, with 6,842 [87.2%] having at least one admission) (Figure Böhm M, et al. 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: Developed by the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) With the special contribution of the Heart Failure Association (HFA) of the ESC.European Heart Journal.2021;42(36):3599-726. 5. Alvarez-Madrazo S, McTaggart S, Nangle C, Nicholson E, Bennie M. Data Resource Profile: The Scottish National Prescribing Information System (PIS).International Journal of Epidemiology.2016;45(3):714-5f.6. Conrad N, Judge A, Tran J, Mohseni H, Hedgecott D, Crespillo AP, et al.Temporal trends and patterns in heart failure incidence: a population-based study of 4 million individuals.Lancet.2018;391(10120):572-80. 7. The Scottish Government.Scottish Index of Multiple Deprivation 2012.In: National Statistics, editor.Edinburgh2012.8. Denaxas SC, George J, Herrett E, Shah AD, Kalra D, Hingorani AD, et al.Data Resource Profile: Cardiovascular disease research using linked bespoke studies and electronic health records (CALIBER).International Journal of Epidemiology.2012;41(6):1625-38.9. Reeves D, Springate DA, Ashcroft DM, Ryan R, Doran T, Morris R, et al.Can analyses of electronic patient records be independently and externally validated?The effect of statins on the mortality of patients Description of the observational cohort for adults with cardiovascular (CV) disease in the Greater Glasgow & Clyde Health Board, including patient characteristics, a breakdown into distinct groups based on loop diuretic (LD) and heart failure (HF) status at baseline, and 5-year mortality statistics.CV disease, neoplasm, infection, and other causes of mortality were reported per patient-year at risk.Cox proportional hazards ratios (HR) and 95% confidence intervals were reported adjusting for age, sex, socioeconomic deprivation, and ten chronic comorbidities.

Figure 1 :
Figure 1: Greater Glasgow & Clyde population classified by sex, age group, repeat prescription

Figure 2 :
Figure 2: Transition diagrams show how many patients started in each of the four groups (left

Figure 3 (
Figure 3 Figure 4: 5-year event rates by patient-year at risk.

Figure 5 :
Figure 5: 5-year survival analysis from 1st January 2012 through end of follow-up classified by Figure 1 559x510 mm ( x DPI) A C C E P T E D M A N U S C R I P T Figure 2 461x559 mm ( x DPI)

Table 1 .
Baseline demographics, co-morbidities, and blood tests for the study population classified by prescription of loop-diuretics 1 (LD) and diagnosis of heart failure (HF) 2 Data are frequencies (%) for categorical values or median (1 st -3 rd quartile) for continuous values.* History of a coded record on or before 1 st January 2012  Defined by the presence of liver fibrosis, sclerosis, or cirrhosis ○ Based on the most recent value in the two years before 1 st January 2012 ◊ Using the World Health Organization's definition of anaemia eGFR = estimated glomerular filtration rate in mL/min/1.73m 2 using CKD-EPI equation 11 Haemoglobin in g/dL H/o hypertension, History of hypertension; CAD, coronary artery disease; MI, myocardial infarction; Data are frequencies (%) for categorical values or median (1 st -3 rd quartile) for continuous values.** A prescription dispensed 180 days before through 1 st January 2012 ‡ Thiazide or thiazide-related medications † either alone or in combination with another agent ACEi, angiotensin-converting-enzyme inhibitors; ARB, angiotensin receptor blockers; MRA, mineralocorticoid receptor antagonists; and NSAIDs, non-steroidal anti-inflammatory drugs.