Abstract
Background
Reliable diagnosis of heart failure during preoperative evaluation is important for perioperative management and long-term care. We aimed to quantify preoperative heart failure diagnostic accuracy and explore characteristics of patients with heart failure misdiagnoses.
Methods
We performed an observational cohort study of adults undergoing major noncardiac surgery at an academic hospital between 2015 and 2019. A preoperative clinical diagnosis of heart failure was defined using keywords from the history and clinical examination or administrative documentation. Across stratified subsamples of cases with and without clinically diagnosed heart failure, health records were intensively reviewed by an expert panel to develop an adjudicated heart failure reference standard using diagnostic criteria congruent with consensus guidelines. We calculated agreement among experts, and analysed performance of clinically diagnosed heart failure compared with the adjudicated reference standard.
Results
Across 40 555 major noncardiac procedures, a stratified subsample of 511 patients was reviewed by the expert panel. The prevalence of heart failure was 9.1% based on clinically diagnosed compared with 13.3% (95% confidence interval [CI], 10.3–16.2%) estimated by the expert panel. Overall agreement and inter-rater reliability (kappa) among heart failure experts were 95% and 0.79, respectively. Based upon expert adjudication, heart failure was clinically diagnosed with an accuracy of 92.8% (90.6–95.1%), sensitivity 57.4% (53.1–61.7%), specificity 98.3% (97.1–99.4%), positive predictive value 83.5% (80.3–86.8%), and negative predictive value 93.8% (91.7–95.9%).
Conclusions
Limitations exist to the preoperative clinical diagnosis of heart failure, with nearly half of cases undiagnosed preoperatively. Considering the risks of undiagnosed heart failure, efforts to improve preoperative heart failure diagnoses are warranted.
Keywords: cardiac risk assessment, diagnostic accuracy, electronic health record, heart failure, noncardiac surgery, observational study, preoperative evaluation
Heart failure is among the greatest risk factors for adverse events after noncardiac surgery, and is independently associated with major complications,1,2 longer postoperative hospital stays,3 more frequent readmissions,4 and higher postoperative mortality.5 Despite advances in heart failure therapies, timely and accurate diagnosis of heart failure remains challenged by the heterogeneity of clinical presentations and course of the disease.6 Among studies of hospitalised patients and outpatients, 25–40% of patients with sufficient electronic health record documentation to define heart failure do not have an established diagnosis of heart failure.7,8 Taken together, these studies suggest that an accurate preoperative diagnosis of heart failure – if leading to improved perioperative management and earlier initiation of guideline-directed medical therapies proven to reduce mortality – potentially carries substantial public health impact for the more than 300 million noncardiac surgical procedures performed annually worldwide.9
During the preoperative surgical evaluation, a wealth of health data (e.g. comprehensive history and clinical examination, laboratory test results, cardiovascular system investigations) are routinely collected, and represent an opportunity for enhanced diagnosis of heart failure. This importance is underscored by findings showing that among patients with heart failure detected by rule-based electronic health record algorithms, a failure to diagnose and document heart failure preoperatively is associated with increased length of stay and mortality.10 Although identifying heart failure preoperatively has the potential to improve outcomes after noncardiac surgery, data are lacking as to the accuracy of heart failure diagnoses by clinicians during preoperative evaluations.
To characterise the accuracy of clinical diagnoses of heart failure in the preoperative setting, we performed an observational cohort study. The aims of this study were to (1) compare the quality of heart failure diagnoses documented preoperatively to those established through expert adjudication; and (2) explore characteristics of patients with heart failure misdiagnoses.
Methods
We obtained institutional review board approval (HUM00143523, University of Michigan, August 8, 2018) for this observational study and patient consent was waived. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines in conducting this study.11 An a priori study protocol was approved within a peer-review forum12 and registered before analysis.13 Data were extracted from the Multicenter Perioperative Outcomes Group (MPOG) electronic anaesthesiology database, our hospital enterprise electronic health record, and a web-based survey tool (Supplementary Methods S1).14, 15, 16 To enable transparency and reproducibility, data were processed using pre-computed, publicly available, universal perioperative electronic health record phenotype algorithms.17
Study design
We conducted an observational analysis of adult patients >40 yr old undergoing index major noncardiac surgical procedures at our quaternary academic medical centre from 1 August 2015 to 31 May 2019. Major noncardiac surgical procedures were defined as those performed under general anaesthesia for >60 min with a Centers for Medicare & Medicaid Services anaesthesiology base unit value >4 (i.e. procedures excluding those with lowest relative value units, such as cataract operations, endoscopies, or skin biopsies). We excluded cardiac surgical procedures as undiagnosed heart failure would be unexpected in this population because of extensive preoperative cardiac evaluation and potential use of intraoperative testing for heart failure (e.g. echocardiography). Additional cases were excluded for similar reasons, and patients with preoperative mechanical circulatory support, inotrope infusions, mechanical ventilation, history of heart or lung transplant, or ASA 5 or 6 physical status classification.
Among cases meeting inclusion criteria, statistically balanced random subsamples of patients with and without a clinical diagnosis of heart failure (described later) were selected for expert review and inclusion in the final analytic dataset. The subset of patients without a preoperative clinical diagnosis of heart failure was further stratified into: (1) high probability patients, defined as those patients lacking a preoperative clinical diagnosis but then developing a clinical diagnosis within 365 days postoperatively and (2) low probability patients, defined as all other patients. To maximise the value of heart failure expert adjudication, patients with a high probability of preoperative heart failure were oversampled; importantly, post-stratification weights were retained in order to determine performance characteristics of preoperative clinical diagnoses of heart failure across the full study cohort. Post-stratification weights were determined by the total number of patients in each subsample: (1) no clinical diagnosis of heart failure/high probability; (2) no clinical diagnosis of heart failure/low probability; and (3) clinical heart failure diagnosis.
Heart failure diagnosis adjudication – expert panel intensive review
To develop a reference standard of patients with and without heart failure, a subset of cases meeting inclusion criteria underwent adjudication via intensive manual review by a clinician panel of heart failure experts (four cardiologists, five cardiac anaesthesiologists, and nine intensivists). All cases were adjudicated by at least two experts; in cases of disagreement, a third expert determined the diagnosis. Before reviews, all experts completed an online training module (Supplementary Methods S2) and underwent calibration on a practice set of patients upon which they received feedback.
To ensure rigorous review before determining an adjudicated heart failure diagnosis, experts completed web-based surveys for determining a heart failure diagnosis (Supplementary Methods S3) with survey time tracked, audited, and attested to by each expert. Experts were required to document all available relevant preoperative cardiac imaging findings (e.g. left ventricular ejection fraction, diastolic function) and all positive/negative mentions (or missingness) of all heart failure signs and symptoms comprising prior established criteria and consensus guidelines for the diagnosis of heart failure.18, 19, 20 Reviewers documented all available diagnostic data for heart failure within the survey, and each reviewer's adjudicated diagnosis of heart failure was based upon their expert judgement in congruence with consensus guidelines and consistent with clinical practice. In addition, reviewers provided their diagnostic certainty (Supplementary Fig. S4).
Heart failure definitions – adjudicated diagnosis vs clinical diagnosis
To maximise diagnostic agreement across adjudicated diagnoses, experts specifically evaluated for American College of Cardiology/American Heart Association (ACC/AHA) guidelines chronic Stage C heart failure (structural heart disease with prior or current symptoms of heart failure) or Stage D heart failure (advanced heart failure). Consistent with guideline recommendations, Stage B heart failure, or structural heart disease in the absence of current or prior symptoms of heart failure, was specifically adjudicated as not heart failure.19 The date of surgery, before the operation, was used as the reference time point for the adjudicated heart failure diagnosis.
Conversely, a preoperative clinical diagnosis of heart failure was defined as either (1) positive mention (structured data or unstructured free text confirmed via manual review) of heart failure (Stage C, D, or unspecified) within the anaesthesia preoperative history and clinical examination, or (2) an International Classification of Diseases (ICD) diagnosis code for heart failure (Supplementary Table S5). Also, we performed a sensitivity analysis in which the clinical diagnosis of heart failure additionally included any patient with a preoperative left ventricular ejection fraction ≤40%, a diagnosis code for cardiomegaly or hypertrophic cardiomyopathy. This analysis was designed to account for patients with ACC/AHA Stage B heart failure, a group with a high likelihood of receiving a heart failure diagnosis preoperatively.
Primary outcome – heart failure clinical diagnosis accuracy
We defined the primary outcome as an accurate preoperative clinical diagnosis of heart failure (true positive or true negative) as compared with the adjudicated heart failure reference standard. Clinical diagnostic accuracy was calculated as (true positive + true negative)/(true positive + true negative + false positive + false negative).
Missing or invalid data
Outlier values were treated as missing if outside of valid ranges described in MPOG phenotype specifications.17 Variables with >10% missing data were excluded from analyses, with the exception of preoperative left ventricular ejection fraction and diastolic dysfunction, which were each classified as categorical variables including ‘missing’.
Statistical analysis
Descriptive statistics were calculated for all perioperative variables, and graphical assessments for normality, symmetry, and potential outliers were performed. Variables showing standardised differences larger than 0.2 in absolute value were considered significant, comparing patients with accurate heart failure diagnoses to those with misdiagnoses. To characterise the validity of the adjudicated heart failure reference standard, the percentage absolute agreement and inter-rater agreement, computed as Cohen's kappa statistic, were used.
Accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of preoperative clinical heart failure diagnoses were calculated using the adjudicated heart failure diagnosis as a reference standard and adjusting for post-stratification weights of each subset reviewed (i.e. number of patients with: (1) a clinical diagnosis of heart failure, (2) no clinical diagnosis of heart failure and high probability patient, and (3) no clinical diagnosis of heart failure and low probability patient). Statistical analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, NC, USA).
Sample size calculation
Among patients adjudicated through expert panel review, for a minimum acceptable kappa of 0.70, expected kappa of 0.80, proportion of adjudicated heart failure diagnoses of 50%, significance level α=0.05 and study power (1–β) of 0.80, a sample of 401 adjudicated patients was required.21 Among patients in the full study cohort, for an expected sensitivity of 75%, specificity of 95%, baseline heart failure prevalence of 5.5%,22 acceptable error of 5.0%, and significance level α=0.05, a sample of 5,239 full study cohort patients was required.23
Results
Among 55 170 adult noncardiac surgical procedures queried, 40 555 met inclusion criteria (Supplementary Fig. S6). Within this full cohort, 3698 cases (9.1%) had a clinical diagnosis of heart failure. Among 36 857 (90.9%) patients without clinical heart failure, 264 (0.7% of full cohort) developed clinical heart failure within 365 days postoperatively (high probability patients), whereas 36 593 (90.2% of full cohort) remained free of clinically diagnosed heart failure (low probability patients). These groups (3698 with clinical heart failure diagnosis; 36 593 with no clinical heart failure diagnosis/low probability; and 264 with no clinical heart failure diagnosis/high probability) determined post-stratification weights as discussed later. Within each group, balanced subsamples of 237 patients with a clinical diagnosis of heart failure and 274 patients without a clinical diagnosis of heart failure (composed of 76 high probability and 198 low probability patients) underwent heart failure expert panel review, totalling 511 patients (Fig. 1).
Fig 1.
Preoperative heart failure (HF) clinical diagnosis and expert adjudication Sankey diagram.
Heart failure expert reviews
During heart failure expert review, median and inter-quartile range (IQR) active review times for each patient were 28 and 19–41 min, respectively. There was agreement among experts (independent of certainty level) for 458 of 511 of patients (90%) with an inter-rater reliability (kappa) of 0.79 (95% confidence interval [CI], 0.74–0.84; Table 1, Table 2). After accounting for post-stratification weights of each subsample used for expert review, estimated reviewer agreement for the full cohort was 95% (94–97%).
Table 1.
Inter-rater agreement of heart failure adjudicated diagnosis among heart failure experts: binary assessment of heart failure.
| Expert X |
|||
|---|---|---|---|
| No heart failure | Heart failure | ||
| Expert Y | No heart failure | 252 (49%) | – |
| Heart failure | 53 (10%) | 206 (40%) | |
Table 2.
Inter-rater agreement of heart failure adjudicated diagnosis among heart failure experts: ordinal assessment of heart failure (HF) with certainty levels.
| Expert X |
|||||||
|---|---|---|---|---|---|---|---|
| Definitely no HF (<5%) | Probably no HF (5–15%) | Possibly no HF (16–49%) | Possibly HF (50–79%) | Probably HF (80–95%) | Definitely HF (>95%) | ||
| Expert Y | Definitely no HF (<5%) | 119 (23.2%) | – | – | – | – | – |
| Probably no HF (5–15%) | 75 (14.7%) | 31 (6.1%) | – | – | – | – | |
| Possibly no HF (16–49%) | 12 (2.3%) | 14 (2.7%) | 1 (0.2%) | – | – | – | |
| Possibly HF (50–79%) | 7 (1.4%) | 16 (3.1%) | 3 (0.6%) | 4 (0.8%) | – | – | |
| Probably HF (80–95%) | 3 (0.6%) | 10 (2.0%) | 1 (0.2%) | 11 (2.2%) | 9 (1.8%) | – | |
| Definitely HF (>95%) | 2 (0.4%) | 8 (1.6%) | 3 (0.6%) | 13 (2.5%) | 32 (6.3%) | 137 (26.8%) | |
Patient baseline characteristics
Baseline characteristics of the full cohort and patients undergoing heart failure expert adjudication are presented in Table 3, Table 4. Data missingness was <10% for all variables in the adjudicated cohort. Overall, the full cohort had a median age of 62 yr (IQR 53–70), 49% were female, and 87% were Caucasian. Among expert-adjudicated patients, the median age was 64 yr (IQR 56–73), 41% were female, and 87% were Caucasian.
Table 3.
Characteristics of the full cohort (n=40 555) and bivariate analyses of patients with and without a clinical diagnosis of heart failure. Data are reported in the form of mean (standard deviation for quantitative variables and n [%] for categorical variables. ∗Classified by Veterans Administration National Drug Formulary Categories. EGFR, estimated glomerular filtration rate; HF, heart failure.
| Variable | Full cohort (n=40,555) | Clinical HF diagnosis (n=3,698) | No clinical HF diagnosis (n=36,857) | Standardised difference |
|---|---|---|---|---|
| Age (yr) | 62 (12) | 68 (120) | 61 (11) | 0.59 |
| Surgical duration (min) | 218 (126) | 223 (126) | 217 (126) | 0.05 |
| Height (cm) | 170 (11) | 171 (11) | 170 (11) | 0.06 |
| Weight (kg) | 87 (23) | 90 (25) | 87 (23) | 0.15 |
| BMI (kg m−2) | 30 (7) | 31 (8) | 30 (7) | 0.13 |
| Baseline MAP (mm Hg) | 97 (13) | 96 (15) | 97 (13) | –0.11 |
| Baseline heart rate (beats min−1) | 75 (14) | 75 (15) | 75 (14) | –0.01 |
| Baseline ventilatory frequency (bpm) | 17 (2) | 17 (2) | 17 (2) | 0.19 |
| Baseline SpO2 (%) | 97 (2) | 96 (2) | 97 (2) | –0.17 |
| Preoperative EGFR (L min−1 1.73 m−2) | 80 (24) | 64 (28) | 81 (23) | –0.68 |
| Preoperative glucose (mg dl−1) | 108 (35) | 118 (43) | 107 (34) | 0.29 |
| Preoperative haemoglobin (g dl−1) | 13.3 (2) | 12.2 (2.2) | 13.4 (1.9) | –0.56 |
| Preoperative platelet count (K μl−1) | 245 (81) | 225 (87) | 247 (80) | –0.26 |
| Preoperative sodium level (mEq L−1) | 140 (3) | 139 (4) | 140 (3) | –0.26 |
| Sex | ||||
| Male | 20667 (51.0) | 2201 (59.5) | 18466 (50.1) | 0.19 |
| Female | 19886 (49.0) | 1497 (40.5) | 18389 (49.9) | |
| Race | ||||
| Asian or Pacific Islander | 903 (2.2) | 63 (1.7) | 840 (2.3) | 0.15 |
| Black, not of Hispanic origin | 2918 (7.2) | 392 (10.6) | 2526 (6.9) | |
| Unknown/Other | 1503 (3.7) | 105 (2.8) | 1398 (3.8) | |
| White, not of Hispanic origin | 35231 (86.9) | 3138 (84.9) | 32093 (87.1) | |
| ASA physical status classification | ||||
| 1 | 765 (1.9) | 2 (0.1) | 763 (2.1) | 1.14 |
| 2 | 15417 (38) | 167 (4.5) | 15250 (41.4) | |
| 3 | 22639 (55.9) | 2758 (74.6) | 19881 (54) | |
| 4 | 1709 (4.2) | 768 (20.8) | 941 (2.6) | |
| Emergent surgery | ||||
| Emergent | 2580 (6.4) | 420 (11.4) | 2160 (5.9) | 0.20 |
| Primary procedural service | ||||
| General | 7253 (17.9) | 550 (14.9) | 6703 (18.2) | 0.52 |
| Otolaryngology | 5666 (14) | 395 (10.7) | 5271 (14.3) | |
| Urology | 4905 (12.1) | 359 (9.7) | 4546 (12.3) | |
| Orthopaedics | 4570 (11.3) | 451 (12.2) | 4119 (11.2) | |
| Neurosurgery | 4124 (10.2) | 278 (7.5) | 3846 (10.4) | |
| Thoracic | 2409 (5.9) | 238 (6.4) | 2171 (5.9) | |
| Trauma | 1953 (4.8) | 253 (6.8) | 1700 (4.6) | |
| Obstetrics/gynaecology | 1707 (4.2) | 91 (2.5) | 1616 (4.4) | |
| Vascular | 1622 (4) | 416 (11.3) | 1206 (3.3) | |
| Plastics | 1596 (3.9) | 59 (1.6) | 1537 (4.2) | |
| Ophthalmology | 1568 (3.9) | 124 (3.4) | 1444 (3.9) | |
| Oral/maxillofacial | 1484 (3.7) | 117 (3.2) | 1367 (3.7) | |
| Transplant | 1067 (2.6) | 201 (5.4) | 866 (2.4) | |
| Dentistry | 334 (0.8) | 20 (0.5) | 314 (0.9) | |
| Other/Unknown | 297 (0.7) | 146 (4) | 151 (0.4) | |
| Comorbidities | ||||
| Hypertension | 22225 (54.8) | 3142 (85) | 19083 (51.8) | 0.76 |
| Cancer | 14643 (36.1) | 1195 (32.3) | 13448 (36.5) | –0.09 |
| Diabetes | 7980 (19.7) | 1442 (39) | 6538 (17.7) | 0.49 |
| Chronic pulmonary disease | 7868 (19.4) | 1240 (33.5) | 6628 (18) | 0.36 |
| Cardiac arrhythmia | 7565 (18.7) | 1856 (50.2) | 5709 (15.5) | 0.80 |
| Hypothyroidism | 5366 (13.2) | 646 (17.5) | 4720 (12.8) | 0.13 |
| Renal failure | 5054 (12.5) | 1436 (38.8) | 3618 (9.8) | 0.72 |
| Coronary artery disease | 4414 (10.9) | 1544 (41.8) | 2870 (7.8) | 0.86 |
| Peripheral vascular disorders | 3194 (7.9) | 1058 (28.6) | 2136 (5.8) | 0.63 |
| Liver disease | 2594 (6.4) | 388 (10.5) | 2206 (6) | 0.16 |
| Coagulopathy | 2054 (5.1) | 476 (12.9) | 1578 (4.3) | 0.31 |
| Valvular disease | 1721 (4.2) | 695 (18.8) | 1026 (2.8) | 0.53 |
| Pulmonary circulation disorders | 1476 (3.6) | 610 (16.5) | 866 (2.4) | 0.5 |
| Cerebrovascular disease | 1374 (3.4) | 295 (8) | 1079 (2.9) | –0.86 |
| Year of surgery | ||||
| 2015 | 4727 (11.7) | 441 (11.9) | 4286 (11.6) | 0.04 |
| 2016 | 10361 (25.6) | 923 (25) | 9438 (25.6) | |
| 2017 | 10541 (26) | 933 (25.2) | 9608 (26.1) | |
| 2018 | 10616 (26.2) | 967 (26.2) | 9649 (26.2) | |
| 2019 | 4310 (10.6) | 434 (11.7) | 3876 (10.5) | |
| Home medications∗ | ||||
| BL110 – Anticoagulants | 3765 (9.3) | 1086 (29.4) | 2679 (7.3) | 0.60 |
| BL117 – Platelet aggregation inhibitors | 1542 (3.8) | 434 (11.7) | 1108 (3) | 0.34 |
| CN101 – Opioids | 4400 (10.9) | 538 (14.6) | 3862 (10.5) | 0.12 |
| CN103 – Non-opioid analgesics | 21490 (53) | 2811 (76) | 18679 (50.7) | 0.54 |
| CV100 – Beta blockers | 11159 (27.5) | 2517 (68.1) | 8642 (23.5) | 1.00 |
| CV150 – Alpha blockers | 3524 (8.7) | 528 (14.3) | 2996 (8.1) | 0.20 |
| CV200 – Calcium channel blockers | 7062 (17.4) | 1007 (27.2) | 6055 (16.4) | 0.26 |
| CV250 – Anti-anginals | 1812 (4.5) | 656 (17.7) | 1156 (3.1) | 0.49 |
| CV300 – Anti-arrhythmics | 2219 (5.5) | 474 (12.8) | 1745 (4.7) | 0.29 |
| CV350 – Anti-lipaemics | 15850 (39.1) | 2344 (63.4) | 13506 (36.6) | 0.56 |
| CV701 – Thiazide diuretics | 6661 (16.4) | 548 (14.8) | 6113 (16.6) | –0.05 |
| CV702 – Loop diuretics | 3216 (7.9) | 1471 (39.8) | 1745 (4.7) | 0.93 |
| CV704 – Potassium-sparing diuretics | 1860 (4.6) | 496 (13.4) | 1364 (3.7) | 0.35 |
| CV800 – ACE inhibitors | 8713 (21.5) | 1121 (30.3) | 7592 (20.6) | 0.22 |
| CV805 – Angiotensin II inhibitors | 4832 (11.9) | 695 (18.8) | 4137 (11.2) | 0.21 |
| HS851 – Thyroid supplements | 6397 (15.8) | 718 (19.4) | 5679 (15.4) | 0.11 |
| RE102 – Inhaled bronchodilators | 6377 (15.7) | 915 (24.7) | 5462 (14.8) | 0.25 |
| Lowest preoperative left ventricular ejection fraction | ||||
| Hyperdynamic (≥70%) | 14 (2.7) | 3 (1.3) | 11 (4) | –1.89 |
| Missing or normal (50–69%) | 290 (56.8) | 54 (22.8) | 236 (86.1) | |
| Mild dysfunction (40–49%) | 46 (9) | 29 (12.2) | 17 (6.2) | |
| Moderate dysfunction (30–39%) | 72 (14.1) | 64 (27) | 8 (2.9) | |
| Severe dysfunction (<30%) | 89 (17.4) | 87 (36.7) | 2 (0.7) | |
| Preoperative diastolic dysfunction | ||||
| Normal or missing | 287 (56.2) | 66 (27.9) | 221 (80.7) | –1.25 |
| Present (any grade) | 224 (43.8) | 171 (72.2) | 53 (19.3) | |
Table 4.
Characteristics of the adjudicated cohort (n=511) and bivariate analyses of patients with and without an adjudicated diagnosis of heart failure (HF). Data are reported in the form of mean (standard deviation for quantitative variables and n (%) for categorical variables. ACE, angiotensin-converting enzyme; EGFR, estimated glomerular filtration rate. ∗Classified by Veterans Administration National Drug Formulary Categories.
| Variable | All adjudicated patients (n=511) | Adjudicated HF diagnosis (n=232) | No adjudicated HF diagnosis (n=279) | Standardised difference |
|---|---|---|---|---|
| Age (yr) | 64 (12) | 67 (11) | 67 (11) | –0.39 |
| Surgical duration (min) | 233 (131) | 230 (120) | 230 (120) | 0.05 |
| Height (cm) | 171 (10) | 172 (10) | 172 (10) | –0.26 |
| Weight (kg) | 84 (23) | 86 (24) | 86 (24) | –0.18 |
| BMI (kg m−2) | 29 (7) | 29 (7) | 29 (7) | –0.08 |
| Baseline MAP (mm Hg) | 94 (15) | 92 (16) | 92 (16) | 0.16 |
| Baseline heart rate (beats min−1) | 76 (15) | 79 (16) | 79 (16) | –0.30 |
| Baseline ventilatory frequency (bpm) | 17 (2) | 17 (3) | 17 (3) | –0.17 |
| Baseline SpO2 (%) | 97 (3) | 96 (3) | 96 (3) | 0.22 |
| Preoperative EGFR (L min−1 1.73 m−2) | 72 (29) | 64 (29) | 64 (29) | 0.52 |
| Preoperative glucose (mg dl−1) | 112 (37) | 119 (44) | 119 (44) | –0.34 |
| Preoperative haemoglobin (g dl−1) | 12.2 (2.4) | 11.7 (2.5) | 11.7 (2.5) | 0.38 |
| Preoperative platelet count (K μl−1) | 234 (103) | 217 (93) | 217 (93) | 0.33 |
| Preoperative sodium level (mEq L−1) | 139 (4) | 139 (4) | 139 (4) | 0.33 |
| Sex | ||||
| Male | 300 (58.7) | 158 (68.1) | 142 (50.9) | 0.36 |
| Female | 211 (41.3) | 74 (31.9) | 137 (49.1) | |
| Race | ||||
| Asian or Pacific Islander | 7 (1.4) | 4 (1.7) | 3 (1.1) | 0.34 |
| Black, not of Hispanic origin | 48 (9.4) | 34 (14.7) | 14 (5.0) | |
| Unknown/Other | 13 (2.5) | 5 (2.2) | 8 (2.9) | |
| White, not of Hispanic origin | 443 (86.7) | 189 (81.5) | 254 (91) | |
| ASA physical status classification | ||||
| 1 | 5 (1.0) | 1 (0.4) | 4 (1.4) | 1.19 |
| 2 | 102 (20.0) | 9 (3.9) | 93 (33.3) | |
| 3 | 304 (59.5) | 133 (57.3) | 171 (61.3) | |
| 4 | 100 (19.6) | 89 (38.4) | 11 (3.9) | |
| Emergent surgery | ||||
| Emergent | 61 (11.9) | 37 (16.0) | 24 (8.6) | 0.23 |
| Primary procedural service | ||||
| General | 69 (13.5) | 25 (10.8) | 44 (15.8) | 0.78 |
| Vascular | 60 (11.7) | 48 (20.7) | 12 (4.3) | |
| Neurosurgery | 53 (10.4) | 11 (4.7) | 42 (15.1) | |
| Orthopaedics | 52 (10.2) | 26 (11.2) | 26 (9.3) | |
| Urology | 46 (9.0) | 17 (7.3) | 29 (10.4) | |
| Otolaryngology | 42 (8.2) | 14 (6.0) | 28 (10.0) | |
| Other/unknown | 42 (8.2) | 25 (10.8) | 17 (6.1) | |
| Trauma | 35 (6.9) | 24 (10.3) | 11 (3.9) | |
| Thoracic | 33 (6.5) | 13 (5.6) | 20 (7.2) | |
| Transplant | 23 (4.5) | 12 (5.2) | 11 (3.9) | |
| Oral/maxillofacial | 16 (3.1) | 5 (2.2) | 11 (3.9) | |
| Obstetrics/gynaecology | 15 (2.9) | 7 (3.0) | 8 (2.9) | |
| Plastics | 13 (2.5) | 2 (0.9) | 11 (3.9) | |
| Ophthalmology | 9 (1.8) | 2 (0.9) | 7 (2.5) | |
| Dentistry | 3 (0.6) | 1 (0.4) | 2 (0.7) | |
| Comorbidities | ||||
| Hypertension | 345 (67.5) | 201 (86.6) | 144 (51.6) | –0.82 |
| Cancer | 157 (30.7) | 63 (27.2) | 94 (33.7) | 0.14 |
| Diabetes | 154 (30.1) | 99 (42.7) | 55 (19.7) | –0.51 |
| Chronic pulmonary disease | 143 (28.0) | 87 (37.5) | 56 (20.1) | –0.39 |
| Cardiac arrhythmia | 235 (46.0) | 143 (61.6) | 92 (33.0) | –0.60 |
| Hypothyroidism | 95 (18.6) | 51 (22.0) | 44 (15.8) | –0.16 |
| Renal failure | 160 (31.3) | 107 (46.1) | 53 (19.0) | –0.60 |
| Coronary artery disease | 180 (35.2) | 130 (56.0) | 50 (17.9) | 0.88 |
| Peripheral vascular disorders | 150 (29.4) | 119 (51.3) | 31 (11.1) | –0.96 |
| Liver disease | 65 (12.7) | 34 (14.7) | 31 (11.1) | –0.11 |
| Coagulopathy | 71 (13.9) | 45 (19.4) | 26 (9.3) | –0.29 |
| Valvular disease | 99 (19.4) | 77 (33.2) | 22 (7.9) | –0.66 |
| Pulmonary circulation disorders | 89 (17.4) | 70 (30.2) | 19 (6.8) | –0.63 |
| Cerebrovascular disease | 37 (7.3) | 24 (10.5) | 13 (4.7) | –0.22 |
| Year of surgery | ||||
| 2015 | 84 (16.4) | 31 (13.4) | 53 (19) | 0.18 |
| 2016 | 169 (33.1) | 78 (33.6) | 91 (32.6) | |
| 2017 | 170 (33.3) | 84 (36.2) | 86 (30.8) | |
| 2018 | 70 (13.7) | 32 (13.8) | 38 (13.6) | |
| 2019 | 18 (3.5) | 7 (3.0) | 11 (3.9) | |
| Home medications∗ | ||||
| BL110 – Anticoagulants | 135 (26.4) | 86 (37.1) | 49 (17.6) | –0.45 |
| BL117 – Platelet aggregation inhibitors | 40 (7.8) | 31 (13.4) | 9 (3.2) | –0.37 |
| CN101 – Opioids | 67 (13.1) | 27 (11.6) | 40 (14.3) | 0.08 |
| CN103 – Non-opioid analgesics | 328 (64.2) | 177 (76.3) | 151 (54.1) | –0.48 |
| CV100 – Beta blockers | 268 (52.4) | 181 (78) | 87 (31.2) | –1.07 |
| CV150 – Alpha blockers | 46 (9.0) | 29 (12.5) | 17 (6.1) | –0.22 |
| CV200 – Calcium channel blockers | 83 (16.2) | 37 (16.0) | 46 (16.5) | 0.01 |
| CV250 – Anti-anginals | 51 (10.0) | 41 (17.7) | 10 (3.6) | –0.47 |
| CV300 – Anti-arrhythmics | 62 (12.1) | 39 (16.8) | 23 (8.2) | –0.26 |
| CV350 – Anti-lipaemics | 246 (48.1) | 140 (60.3) | 106 (38.0) | –0.46 |
| CV701 – Thiazide diuretics | 51 (10.0) | 22 (9.5) | 29 (10.4) | 0.03 |
| CV702 – Loop diuretics | 123 (24.1) | 104 (44.8) | 19 (6.8) | –0.96 |
| CV704 – Potassium-sparing diuretics | 57 (11.2) | 47 (20.3) | 10 (3.6) | –0.53 |
| CV800 – ACE inhibitors | 130 (25.4) | 88 (37.9) | 42 (15.1) | –0.54 |
| CV805 – Angiotensin II inhibitors | 65 (12.7) | 32 (13.8) | 33 (11.8) | –0.06 |
| HS851 – Thyroid supplements | 94 (18.4) | 46 (19.8) | 48 (17.2) | –0.07 |
| RE102 – Inhaled bronchodilators | 86 (16.8) | 45 (19.4) | 41 (14.7) | –0.13 |
| Lowest preoperative left ventricular ejection fraction | ||||
| Hyperdynamic (≥70%) | 14 (2.7) | 12 (4.3) | 2.27 | |
| Missing or normal (50–69%) | 290 (56.8) | 248 (88.9) | ||
| Mild dysfunction (40–49%) | 46 (9) | 9 (3.2) | ||
| Moderate dysfunction (30–39%) | 72 (14.1) | 5 (1.8) | ||
| Severe dysfunction (<30%) | 89 (17.4) | 5 (1.8) | ||
| Preoperative diastolic dysfunction | ||||
| Normal or missing | 287 (56.2) | 225 (80.7) | –1.29 | |
| Present (any grade) | 224 (43.8) | 54 (19.4) | ||
Study outcomes – performance of clinically diagnosed heart failure
After expert review, 39 of 237 cases with a clinical heart failure diagnosis (16%) were determined not to have heart failure (false positives), and 34 of 274 cases without preoperative heart failure documentation (12%) were determined to have heart failure (false negatives) (Table 5). After accounting for post-stratification weights and the baseline prevalence of clinically diagnosed heart failure (9.1%), the true prevalence of heart failure across the overall surgical cohort was estimated to be significantly higher at 13.3% (95% CI, 10.3–16.2%).
Table 5.
Preoperative clinical diagnosis vs adjudicated diagnosis of heart failure among patients undergoing expert review (n=511).
| Clinical diagnosis positive (n=237) | Clinical diagnosis negative (n=274) | |
|---|---|---|
| Adjudicated diagnosis positive (n=232) | 198 (True positive) | 34 (False negative) |
| Adjudicated diagnosis negative (n=279) | 39 (False positive) | 240 (True negative) |
Using the adjudicated diagnosis of heart failure as a reference standard and adjusting for post-stratification weights of subsets reviewed, the estimated accuracy of the preoperative clinical diagnosis of heart failure was 92.8% (95% CI, 90.6–95.1%). In addition, the estimated sensitivity of clinically diagnosed heart failure was 57.4% (53.1–61.7%), specificity 98.3% (97.1–99.4%), positive predictive value 83.5% (80.3–86.8%), and negative predictive value 93.8% (91.7–95.9%).
The 13.3% of patients with an adjudicated diagnosis of heart failure (true positives + false negatives) was composed of 7.6% (5.3–9.9%) with a clinical diagnosis (true positives) and 5.7% (3.7–7.7%) without a clinical diagnosis (false negatives). Thus, almost half (i.e. 42.6% [38.3–46.9%]) of patients with heart failure preoperatively were undiagnosed by clinicians (1–sensitivity). Compared with patients with a clinical diagnosis of heart failure (true positives), those without a clinical diagnosis (false negatives) were more commonly younger and female; had higher left ventricular ejection fractions, less diastolic dysfunction, fewer comorbidities, lower BMIs, lower haemoglobin A1c concentrations and international normalised ratio coagulation assays, higher platelet counts, and lower ASA physical status classifications; and more frequently underwent trauma surgery (Table 6).
Table 6.
Characteristics of patients with correct vs incorrect clinical heart failure diagnoses as compared with adjudicated heart failure reference standard. Data are reported in the form of mean (standard deviation for quantitative variables and n [%] for categorical variables. ACE, angiotensin-converting enzyme; EGFR, estimated glomerular filtration rate; HF, heart failure. ∗Classified by Veterans Administration National Drug Formulary Categories.
| Patients with an adjudicated diagnosis of heart failure (True positives + false negatives) |
Patients without an adjudicated diagnosis of heart failure (True negatives + false positives) |
|||||
|---|---|---|---|---|---|---|
| Variable | Patients with a clinical diagnosis of heart failure ‘True positives’ (n=198) |
Patients without a clinical diagnosis of heart failure ‘False negatives’ (n=34) |
Standardised difference | Patients without a clinical diagnosis of heart failure ‘True negatives’ (n=240) |
Patients with a clinical diagnosis of heart failure ‘False positives’ (n=39) |
Standardised difference |
| Age (yr) | 67 (11) | 65 (11) | –0.22 | 61 (12) | 70 (11) | 0.76 |
| Surgical duration (min) | 230 (118) | 228 (135) | –0.02 | 237 (143) | 233 (118) | –0.03 |
| Height (cm) | 173 (10) | 169 (10) | –0.44 | 170 (10) | 170 (12) | 0.01 |
| Weight (kg) | 87 (23) | 84 (27) | –0.12 | 83 (22) | 80 (25) | –0.10 |
| BMI (kg m−2) | 29 (7) | 29 (8) | 0.05 | 29 (7) | 28 (8) | –0.12 |
| Baseline MAP (mm Hg) | 91 (15) | 101 (14) | 0.67 | 95 (15) | 93 (12) | –0.17 |
| Baseline heart rate (beats min−1) | 79 (16) | 81 (14) | 0.17 | 74 (14) | 78 (18) | 0.28 |
| Baseline ventilatory frequency (bpm) | 17 (2) | 17 (3) | –0.14 | 17 (2) | 17 (2) | 0.23 |
| Baseline SpO2 (%) | 96 (3) | 96 (2) | –0.04 | 97 (2) | 97 (3) | –0.17 |
| Preoperative EGFR (L min−1 1.73 m−2) | 64 (29) | 68 (32) | 0.15 | 81 (26) | 67 (32) | –0.51 |
| Preoperative glucose (mg dl−1) | 120 (45) | 113 (34) | –0.17 | 106 (29) | 107 (32) | 0.06 |
| Preoperative haemoglobin (g dl−1) | 12 (2) | 13 (3) | 0.38 | 13 (2) | 12 (3) | –0.40 |
| Preoperative platelet count (K μl−1) | 209 (86) | 264 (117) | 0.53 | 249 (99) | 260 (156) | 0.09 |
| Preoperative sodium level (mEq L−1) | 139 (4) | 138 (4) | –0.05 | 140 (3) | 140 (4) | –0.08 |
| Sex | ||||||
| Male | 138 (69.7) | 20 (58.8) | 0.23 | 120 (50.0) | 22 (56.4) | –0.13 |
| Female | 60 (30.3) | 14 (41.2) | 120 (50.0) | 17 (43.6) | ||
| Race | ||||||
| Asian or Pacific Islander | 4 (2.0) | 0 (0) | 0.31 | 2 (0.8) | 1 (2.6) | 0.29 |
| Black, not of Hispanic origin | 29 (14.7) | 5 (14.7) | 12 (5) | 2 (5.1) | ||
| Unknown/other | 3 (1.5) | 2 (5.9) | 8 (3.3) | (0) | ||
| White, not of Hispanic origin | 162 (81.8) | 27 (79.4) | 218 (90.8) | 36 (92.3) | ||
| ASA physical status classification | ||||||
| 1 | 1 (0.5) | 0 (0) | 0.56 | 4 (1.7) | 0 (0) | 1.11 |
| 2 | 5 (2.5) | 4 (11.8) | 92 (38.3) | 1 (2.6) | ||
| 3 | 110 (55.6) | 23 (67.7) | 139 (57.9) | 32 (82.1) | ||
| 4 | 82 (41.4) | 7 (20.6) | 5 (2.1) | 6 (15.4) | ||
| Emergent surgery | ||||||
| Emergent | 30 (15.2) | 7 (20.6) | 0.14 | 17 (7.1) | 7 (18.0) | 0.33 |
| Primary procedural service | ||||||
| General | 22 (11.1) | 3 (8.8) | 0.87 | 37 (15.4) | 7 (18.0) | 1.24 |
| Vascular | 43 (21.7) | 5 (14.7) | 9 (3.8) | 3 (7.7) | ||
| Neurosurgery | 8 (4.0) | 3 (8.8) | 37 (15.4) | 5 (12.8) | ||
| Orthopaedics | 23 (11.6) | 3 (8.8) | 22 (9.2) | 4 (10.3) | ||
| Urology | 14 (7.1) | 3 (8.8) | 29 (12.1) | 0 (0) | ||
| Otolaryngology | 12 (6.1) | 2 (5.9) | 28 (11.7) | 0 (0) | ||
| Other/unknown | 25 (12.6) | 0 (0) | 9 (3.8) | 8 (20.5) | ||
| Trauma | 17 (8.6) | 7 (20.6) | 8 (3.3) | 3 (7.7) | ||
| Thoracic | 11 (5.6) | 2 (5.9) | 14 (5.8) | 6 (15.4) | ||
| Transplant | 11 (5.6) | 1 (2.9) | 9 (3.8) | 2 (5.1) | ||
| Oral/maxillofacial | 2 (1.0) | 3 (8.8) | 11 (4.6) | 0 (0) | ||
| Obstetrics/gynaecology | 6 (3.0) | 1 (2.9) | 8 (3.3) | 0 (0) | ||
| Plastics | 1 (0.5) | 1 (2.9) | 10 (4.2) | 1 (2.6) | ||
| Ophthalmology | 2 (1.0) | 0 (0) | 7 (2.9) | 0 (0) | ||
| Dentistry | 1 (0.5) | 0 (0) | 2 (0.8) | 0 (0) | ||
| Comorbidities | ||||||
| Hypertension | 174 (87.9) | 27 (79.4) | –0.23 | 111 (46.3) | 33 (84.6) | 0.88 |
| Cancer | 55 (27.8) | 8 (23.5) | –0.10 | 85 (35.4) | 9 (23.1) | –0.27 |
| Diabetes | 89 (45.0) | 10 (29.4) | –0.33 | 48 (20.0) | 7 (18.0) | –0.05 |
| Chronic pulmonary disease | 74 (37.4) | 13 (38.2) | 0.02 | 44 (18.3) | 12 (30.8) | 0.29 |
| Cardiac arrhythmia | 126 (63.6) | 17 (50.0) | –0.28 | 61 (25.4) | 31 (79.5) | 1.29 |
| Hypothyroidism | 42 (21.2) | 9 (26.5) | 0.12 | 35 (14.6) | 9 (23.1) | 0.22 |
| Renal failure | 95 (48.0) | 12 (35.3) | –0.26 | 40 (16.7) | 13 (33.3) | 0.39 |
| Coronary artery disease | 115 (58.1) | 15 (44.1) | 0.35 | 33 (13.8) | 17 (43.6) | 0.77 |
| Peripheral vascular disorders | 108 (54.6) | 11 (32.4) | –0.46 | 22 (9.2) | 9 (23.1) | 0.39 |
| Liver disease | 30 (15.2) | 4 (11.8) | –0.10 | 27 (11.3) | 4 (10.3) | –0.03 |
| Coagulopathy | 42 (21.2) | 3 (8.8) | –0.35 | 15 (6.3) | 11 (28.2) | 0.61 |
| Valvular disease | 70 (35.4) | 7 (20.6) | –0.33 | 13 (5.4) | 9 (23.1) | 0.52 |
| Pulmonary circulation disorders | 64 (32.3) | 6 (17.7) | –0.34 | 15 (6.3) | 4 (10.3) | 0.15 |
| Cerebrovascular disease | 20 (10.3) | 4 (11.8) | 0.05 | 9 (3.8) | 4 (10.5) | 0.27 |
| Year of surgery | ||||||
| 2015 | 25 (12.6) | 6 (17.7) | 0.22 | 46 (19.2) | 7 (18) | 0.19 |
| 2016 | 67 (33.8) | 11 (32.4) | 76 (31.7) | 15 (38.5) | ||
| 2017 | 74 (37.4) | 10 (29.4) | 76 (31.7) | 10 (25.6) | ||
| 2018 | 26 (13.1) | 6 (17.7) | 32 (13.3) | 6 (15.4) | ||
| 2019 | 6 (3.0) | 1 (2.9) | 10 (4.2) | 1 (2.6) | ||
| Home medications∗ | ||||||
| BL110 – Anticoagulants | 80 (40.4) | 6 (17.7) | –0.52 | 30 (12.5) | 19 (48.7) | 0.85 |
| BL117 – Platelet aggregation inhibitors | 29 (14.7) | 2 (5.9) | –0.29 | 7 (2.9) | 2 (5.1) | 0.11 |
| CN101 – Opioids | 26 (13.1) | 1 (2.9) | –0.38 | 38 (15.8) | 2 (5.1) | –0.36 |
| CN103 – Non-opioid analgesics | 157 (79.3) | 20 (58.8) | –0.45 | 122 (50.8) | 29 (74.4) | 0.50 |
| CV100 – Beta blockers | 159 (80.3) | 22 (64.7) | –0.35 | 63 (26.3) | 24 (61.5) | 0.76 |
| CV150 – Alpha blockers | 25 (12.6) | 4 (11.8) | –0.03 | 13 (5.4) | 4 (10.3) | 0.18 |
| CV200 – Calcium channel blockers | 30 (15.2) | 7 (20.6) | 0.14 | 32 (13.3) | 14 (35.9) | 0.54 |
| CV250 – Anti-anginals | 36 (18.2) | 5 (14.7) | –0.09 | 8 (3.3) | 2 (5.1) | 0.09 |
| CV300 – Anti-arrhythmics | 35 (17.7) | 4 (11.8) | –0.17 | 18 (7.5) | 5 (12.8) | 0.18 |
| CV350 – Anti-lipaemics | 121 (61.1) | 19 (55.9) | –0.11 | 87 (36.3) | 19 (48.7) | 0.25 |
| CV701 – Thiazide diuretics | 17 (8.6) | 5 (14.7) | 0.19 | 23 (9.6) | 6 (15.4) | 0.18 |
| CV702 – Loop diuretics | 98 (49.5) | 6 (17.7) | –0.72 | 12 (5) | 7 (18) | 0.41 |
| CV704 – Potassium-sparing diuretics | 46 (23.2) | 1 (2.9) | –0.63 | 8 (3.3) | 2 (5.1) | 0.09 |
| CV800 – ACE inhibitors | 75 (37.9) | 13 (38.2) | 0.01 | 34 (14.2) | 8 (20.5) | 0.17 |
| CV805 – Angiotensin II inhibitors | 29 (14.7) | 3 (8.8) | –0.18 | 25 (10.4) | 8 (20.5) | 0.28 |
| HS851 – Thyroid supplements | 36 (18.2) | 10 (29.4) | 0.27 | 41 (17.1) | 7 (18) | 0.02 |
| RE102 – Inhaled bronchodilators | 39 (19.7) | 6 (17.7) | –0.05 | 34 (14.2) | 7 (18) | 0.10 |
| Lowest preoperative left ventricular ejection fraction | ||||||
| Hyperdynamic (≥70%) | 1 (0.5) | 1 (2.9) | 1.14 | 10 (4.2) | 2 (5.1) | 0.84 |
| Missing or normal (50–69%) | 29 (14.7) | 13 (38.2) | 223 (92.9) | 25 (64.1) | ||
| Mild dysfunction (40–49%) | 26 (13.1) | 11 (32.4) | 6 (2.5) | 3 (7.7) | ||
| Moderate dysfunction (30–39%) | 60 (30.3) | 7 (20.6) | 1 (0.4) | 4 (10.3) | ||
| Severe dysfunction (<30%) | 82 (41.4) | 2 (5.9) | 0 (0) | 5 (12.8) | ||
| Preoperative diastolic dysfunction | ||||||
| Normal or missing | 41 (20.7) | 21 (61.8) | –0.92 | 200 (83.3) | 25 (64.1) | 0.45 |
| Present (any grade) | 157 (79.3) | 13 (38.2) | 40 (16.7) | 14 (35.9) | ||
Conversely, the estimated 86.7% of patients without an adjudicated diagnosis of heart failure (true negatives + false positives) was composed of 85.2% (82.1–88.3%) without a clinical diagnosis (true negatives) and 1.5% (0.4–2.6%) with a clinical diagnosis (false positives). This corresponded to 1.7% (0.6–2.9%) of patients being incorrectly diagnosed as having heart failure by clinicians. Compared with patients without a clinical diagnosis of heart failure (true negatives), patients with a clinical diagnosis (false positives) were older; had more cardiovascular comorbidities, lower left ventricular ejection fractions, less diastolic dysfunction, higher BMIs, higher ASA physical status classifications; and more frequently underwent thoracic, trauma, or vascular surgery.
Sensitivity analysis
In a sensitivity analysis in which clinically diagnosed heart failure was expanded to include patients with a preoperative left ventricular ejection fraction ≤40%, irrespective of the presence of a heart failure diagnosis, or had a diagnosis code for cardiomegaly or hypertrophic cardiomyopathy, 315 additional patients were identified (7.8% of total patients with clinically diagnosed heart failure, n=4013). Heart failure diagnostic accuracy improved to 94.4% (95% CI, 92.4–96.4%), sensitivity to 67.7% (63.7–71.8%), and negative predictive value to 95.7% (93.9–97.4%). Conversely, a decrease was observed in specificity 98.0% (96.8–99.2%) and positive predictive value 82.6% (79.3–85.9%). Characteristics of patients with accurate diagnoses vs misdiagnoses related to heart failure were similar to the primary analysis.
Discussion
To understand the accuracy of heart failure diagnosed clinically during the preoperative surgical evaluation, we performed this observational cohort study which used a panel of heart failure experts to perform intensive chart reviews of older adults undergoing major noncardiac surgeries. We report five major findings.
First, the estimated true baseline prevalence of adjudicated heart failure in patients presenting for major noncardiac surgery under general anaesthesia was 13.3% based upon an expert panel review with high diagnostic agreement (95%). Compared with previous studies examining heart failure in surgical populations,22,24 this prevalence of heart failure in our cohort was substantially higher. The higher prevalence was likely attributable to (1) our study inclusion criteria and large academic medical centre setting, favouring older patients with more comorbid conditions undergoing major non-outpatient surgeries; and (2) shortcomings to clinical diagnostic sensitivity of heart failure in previous studies, as later discussed.
Second, the prevalence of preoperative clinical diagnoses of heart failure (9.1%) underestimated the true baseline prevalence. Under-recognition of heart failure highlights that clinical diagnoses primarily lack diagnostic sensitivity rather than specificity; this finding is consistent with previous literature.8,22,25 Based on the clinical diagnostic sensitivity of heart failure in our study, nearly half of patients with adjudicated heart failure were missed during their preoperative evaluation. Given that heart failure remains one of the most significant risk factors for morbidity and mortality after noncardiac surgery1 2 5 and leads to substantially increased healthcare costs3 and readmissions,4 our findings highlight missed opportunities for early recognition, preoperative optimisation, and surgical risk reduction (e.g. avoidance of volume overload,26 additional haemodynamic monitoring,27 and anaesthetic medication adjustments28,29) among patients with undiagnosed heart failure. Furthermore, early diagnosis has the potential to improve the longitudinal health trajectories of patients with heart failure, irrespective of short-term surgical outcomes, through timely initiation of guideline-directed medical therapy.
Third, compared with ‘true positive’ patients with both a clinical and adjudicated diagnosis of heart failure, ‘false negative’ patients without a clinical diagnosis of heart failure, yet with an adjudicated diagnosis, had fewer markers of poor health and were more likely to be female. The lack of markers for poor health was not likely attributable to incomplete medical documentation (e.g. failure to document other comorbidities), as we also observed this trend for electronic health record characteristics that were collected and recorded in an automated fashion (e.g. routine preoperative laboratory values). Rather, differences between ‘false negative’ and ‘true positive’ patients were potentially explained by a lower index of clinical suspicion for heart failure in these patients. The increased likelihood of being female may be explained by ‘false negative’ patients tending to be younger, with heart failure known to develop at a later age in females30; however, additional under-recognised inequities in heart failure diagnosis31 may also explain this finding.
Fourth, compared with ‘true negative’ patients who lacked both a clinical and adjudicated diagnosis of heart failure, ‘false positive’ patients with a clinical diagnosis of heart failure, yet without an adjudicated diagnosis, more commonly had additional markers of poor health. Similar to the previous finding, this may be explained by the association between these negative health markers and heart failure, raising clinical suspicion for the disease.
Finally, whereas the overall diagnostic accuracy for clinical heart failure was high (92.8%), the sensitivity (57.4%) was lower than that of previous studies which reported sensitivities ranging from 70% to 90%.8 25 This difference, which was also observed in our sensitivity analysis, was likely not attributable to limitations in ascertainment of clinical diagnoses within the electronic health record. To the contrary, our clinical diagnosis definition included diagnosis codes and keywords within the preoperative history and clinical examination which biased towards greater sensitivity compared with prior studies restricted to administrative data. Rather, the difference may be explained by the rigour of chart review through an expert consensus-adjudicated reference standard with reviewer training, calibration, and auditing. Such findings may have important implications for perioperative epidemiological studies and prediction models not using expert adjudication and therefore relying on complete and accurate heart failure clinical documentation.32, 33, 34
Study limitations
Our study has several important limitations. First, the study was performed at a single academic medical centre among primarily Caucasian patients. Although the full cohort included a large population across a wide range of surgical procedures with validated variables, the cohort adjudicated by heart failure experts focused on a relatively smaller number of patients. This trade-off between data quantity and quality favoured the lower number of high-quality reviews, potentially offering unique insights compared with larger studies using less well-defined schema for identifying patients with heart failure. Second, the study was observational in nature. As such, heart failure expert reviewers only had access to the electronic health record data, rather than an in-person evaluation with each patient reviewed. Such limitations were mitigated through the use of a consensus panel of two or three experts rather than a single reviewer, and an ability to review electronic health record data up to 365 days after surgery (with sequelae such as prolonged hospitalisations or readmissions occasionally influencing an expert's adjudicated preoperative diagnosis of heart failure). This limitation was further explored through quantification of expert diagnostic certainty. Third, whereas the study defined a clinical diagnosis of heart failure from multiple data sources, the diagnosis relied upon electronic health record documentation and did not necessarily equate to the perioperative care team's awareness of heart failure. Nevertheless, failure to document a clinical diagnosis of heart failure by the perioperative care team remains an important finding, given the increased risk of postoperative complications,10 and may have downstream consequences for clinicians later involved in the care of such patients.
Conclusions and next steps
We describe a heart failure clinical diagnostic accuracy of 92.8% for older patients undergoing major noncardiac surgical procedures at a single academic medical centre. Among the 13.3% of patients in this cohort who were projected to have heart failure by the expert panel, almost one-half of diagnoses were missed during preoperative evaluation. Given the substantial health risks posed by undiagnosed heart failure on postoperative outcomes and long-term health trajectories, our findings may represent a call to action for improved preoperative clinical diagnosis of heart failure.
To determine whether improved preoperative clinical diagnoses of heart failure may lead to improved perioperative care, postoperative outcomes, and long-term patient health trajectories, several future studies may be pursued as next steps. These include studies exploring potential associations between heart failure misdiagnoses and heart failure-related intraoperative practice patterns such as fluid balance, haemodynamic management, anaesthetic techniques, and invasive monitoring; and similar studies exploring postoperative outcomes such as complications (e.g. acute kidney injury, pulmonary complications), hospital length of stay, and heart failure-related readmissions. Should differences in intraoperative heart failure-related practice patterns and postoperative outcomes be observed among patients with heart failure misdiagnoses, subsequent prospective interventional studies seeking to reduce preoperative misdiagnosis of heart failure are warranted. These may include studies which explore the impact of electronic health record-based preoperative screening algorithms for heart failure, with an emphasis on ‘false-negative’ and ‘false-positive’ patients identified in this study, and studies which explore the impact of goal-directed heart failure-related perioperative management strategies among commonly misdiagnosed patients.
Authors' contributions
Concept: MRM
Design of the work: JRG, HJ, MRM
Analyses performed: JRG
Statistical analyses: RBC
Querying and curation of the cohort studied: HJ
All authors were involved in the interpretation of data, revisions to the work for important intellectual content, and final approval of the version to be published, and agreed to accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Funding
US National Institutes of Health – National Heart Lung and Blood Institute (Bethesda, MD, USA) (K01 HL141701); the Department of Anesthesiology, Michigan Medicine, University of Michigan (Ann Arbor, MI, USA). The opinions, beliefs, and viewpoints expressed by the authors do not necessarily reflect the opinions, beliefs, and viewpoints of the National Institutes of Health, or any of its employees. Industry contributors have had no role in the study.
Declaration of interest
The authors declare: JRG received funding from the US National Institutes of Health (L30HL143700) and received salary support by an American Heart Association grant (20SFRN35370008). MRM has received a research grant from the US National Institutes of Health (NHLBI K01HL141701). AMJ has received a research grant from the US National Institutes of Health (NIGMS T32GM103730), has received research support paid to the University of Michigan, and unrelated to this present work, from Becton, Dickinson, and Company. SK reported receiving grants from the US National Institutes of Health (NIH), Blue Cross Blue Shield of Michigan (BCBSM), Merck Inc., Apple Inc., and Becton Dickinson Inc. outside the submitted work.
Acknowledgements
The authors gratefully acknowledge Robert Coleman (Department of Anesthesiology, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA) for contributions in data acquisition and electronic search query programming for this project. They also thank Supriya Shore (Department of Internal Medicine, Division of Cardiovascular Medicine, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA); Sina Jame (Department of Internal Medicine, Division of Cardiovascular Medicine, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA); Marty Tam, (Department of Internal Medicine, Division of Cardiovascular Medicine, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA); Sharaf Khan (Department of Emergency Medicine, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA); Mark Korenke (Department of Anesthesiology, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA); Matthew Polasko (Department of Anesthesiology, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA); and Daniel Horner (Department of Anesthesiology, Michigan Medicine, University of Michigan, Ann Arbor, MI, USA) for contributions in performing electronic health record manual reviews (in addition to co-authors) to ascertain heart failure adjudicated diagnoses as clinical experts.
Handling editor: Phil Hopkins
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.bjao.2022.100113.
Contributor Information
Michael R. Mathis, Email: mathism@med.umich.edu.
Michigan Congestive Heart Failure Investigators:
Graciela B. Mentz, Brahmajee K. Nallamothu, Francis D. Pagani, Donald S. Likosky, and Thomas M. Cascino
Appendix 1
Group non-author collaborators – Michigan Congestive Heart Failure Investigators (PubMed-indexed): Graciela B. Mentz, Senior statistician; Brahmajee K. Nallamothu, Professor; Francis D. Pagani, Professor; Donald S. Likosky, Professor; Thomas M. Cascino, Clinical instructor.
Appendix A. Supplementary data
The following are the Supplementary data to this article:
References
- 1.Duceppe E., Patel A., Chan M.T.V., et al. Preoperative N-terminal pro-B-type natriuretic peptide and cardiovascular events after noncardiac surgery: a cohort study. Ann Intern Med. 2020;172:96–104. doi: 10.7326/M19-2501. [DOI] [PubMed] [Google Scholar]
- 2.Smit-Fun V., Buhre W.F. The patient with chronic heart failure undergoing surgery. Curr Opin Anaesthesiol. 2016;29:391–396. doi: 10.1097/ACO.0000000000000335. [DOI] [PubMed] [Google Scholar]
- 3.Xu-Cai Y.O., Brotman D.J., Phillips C.O., et al. Outcomes of patients with stable heart failure undergoing elective noncardiac surgery. Mayo Clin Proc. 2008;83:280–288. doi: 10.4065/83.3.280. [DOI] [PubMed] [Google Scholar]
- 4.van Diepen S., Bakal J.A., McAlister F.A., Ezekowitz J.A. Mortality and readmission of patients with heart failure, atrial fibrillation, or coronary artery disease undergoing noncardiac surgery: an analysis of 38 047 patients. Circulation. 2011;124:289–296. doi: 10.1161/CIRCULATIONAHA.110.011130. [DOI] [PubMed] [Google Scholar]
- 5.Fleisher L.A., Fleischmann K.E., Auerbach A.D., et al. ACC/AHA guideline on perioperative cardiovascular evaluation and management of patients undergoing noncardiac surgery: a report of the American College of Cardiology/American Heart Association Task Force on practice guidelines. J Am Coll Cardiol. 2014;64:e77–e137. doi: 10.1016/j.jacc.2014.07.944. [DOI] [PubMed] [Google Scholar]
- 6.Pakhomov S., Weston S.A., Jacobsen S.J., Chute C.G., Meverden R., Roger V.L. Electronic medical records for clinical research: application to the identification of heart failure. Am J Manag Care. 2007;13:281–288. [PubMed] [Google Scholar]
- 7.Rosamond W.D., Chang P.P., Baggett C., et al. Classification of heart failure in the atherosclerosis risk in communities (ARIC) study: a comparison of the diagnostic criteria. Circ Heart Fail. 2012;5:152–159. doi: 10.1161/CIRCHEARTFAILURE.111.963199. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.McCormick N., Lacaille D., Bhole V., Avina-Zubieta J.A. Validity of heart failure diagnoses in administrative databases: a systematic review and meta-analysis. PLoS One. 2014;9 doi: 10.1371/journal.pone.0104519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Weiser T.G., Haynes A.B., Molina G., et al. Estimate of the global volume of surgery in 2012: an assessment supporting improved health outcomes. Lancet. 2015;385:S11. doi: 10.1016/S0140-6736(15)60806-6. [DOI] [PubMed] [Google Scholar]
- 10.Hofer I.S., Cheng D., Grogan T. A retrospective analysis demonstrates that a failure to document key comorbid diseases in the anesthesia preoperative evaluation associates with increased length of stay and mortality. Anesth Analg. 2021;133:698–706. doi: 10.1213/ANE.0000000000005393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Benchimol E.I., Smeeth L., Guttmann A., et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med. 2015;12 doi: 10.1371/journal.pmed.1001885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Anesthesia Clinical Research Committee . University of Michigan Anesthesiology Clinical Research; 2022. https://anes.med.umich.edu/research/acrc.html accessed. [Google Scholar]
- 13.Mathis M.R., Singh K., Joo H., et al. Preoperative diagnostic accuracy of heart failure among patients undergoing major noncardiac surgery. Anesth Analg. 2022;134(suppl):S95–S97. doi: 10.1016/j.bjao.2022.100113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Sun E., Mello M.M., Rishel C.A., et al. Association of overlapping surgery with perioperative outcomes. JAMA. 2019;321:762–772. doi: 10.1001/jama.2019.0711. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Burns M.L., Saager L., Cassidy R.B., Mentz G., Mashour G.A., Kheterpal S. Association of anesthesiologist staffing ratio with surgical patient morbidity and mortality. JAMA Surg. 2022;157:807–815. doi: 10.1001/jamasurg.2022.2804. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Colquhoun D.A., Shanks A.M., Kapeles S.R., et al. Considerations for integration of perioperative electronic health records across institutions for research and quality improvement: the approach taken by the Multicenter Perioperative Outcomes Group. Anesth Analg. 2020;130:1133–1146. doi: 10.1213/ANE.0000000000004489. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.MPOG phenotype browser. Multicenter perioperative outcomes group. https://collations.mpogresearch.org/(accessed 21 April 2022).
- 18.McKee P.A., Castelli W.P., McNamara P.M., Kannel W.B. The natural history of congestive heart failure: the Framingham study. N Engl J Med. 1971;285:1441–1446. doi: 10.1056/NEJM197112232852601. [DOI] [PubMed] [Google Scholar]
- 19.Heidenreich P.A., Bozkurt B., Aguilar D., et al. vol. 79. J Am Coll Cardiol Elsevier BV; 2022. (AHA/ACC/HFSA guideline for the management of heart failure: a report of the American College of Cardiology/American Heart Association joint committee on clinical practice guidelines). e263–421. [DOI] [PubMed] [Google Scholar]
- 20.Ponikowski P., Voors A.A., Anker S.D., et al. ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur Heart J. 2016;37:2129–2200. doi: 10.1093/eurheartj/ehw128. [DOI] [PubMed] [Google Scholar]
- 21.Arifin W.N. 2017. Sample size calculator.http://wnarifin.github.io accessed. [Google Scholar]
- 22.Lerman B.J., Popat R.A., Assimes T.L., Heidenreich P.A., Wren S.M. Association between heart failure and postoperative mortality among patients undergoing ambulatory noncardiac surgery. JAMA Surg. 2019;154:907–914. doi: 10.1001/jamasurg.2019.2110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Buderer N.M.F., Fenn Buderer N.M. Statistical methodology: I. Incorporating the prevalence of disease into the sample size calculation for sensitivity and specificity. Acad Emerg Med. 1996;3:895–900. doi: 10.1111/j.1553-2712.1996.tb03538.x. [DOI] [PubMed] [Google Scholar]
- 24.Mattingly A.S., Lerman B.J., Popat R., Wren S.M. Association of sex with postoperative mortality among patients with heart failure who underwent elective noncardiac operations. JAMA Netw Open. 2019;2 doi: 10.1001/jamanetworkopen.2019.14420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Floyd J.S., Blondon M., Moore K.P., Boyko E.J., Smith N.L. Validation of methods for assessing cardiovascular disease using electronic health data in a cohort of Veterans with diabetes. Pharmacoepidemiol Drug Saf. 2016;25:467–471. doi: 10.1002/pds.3921. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Navarro L.H.C., Bloomstone J.A., Auler J.O.C., Jr., et al. Perioperative fluid therapy: a statement from the international Fluid Optimization Group. Perioper Med (Lond) 2015;4:3. doi: 10.1186/s13741-015-0014-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Vincent J.-L., Pelosi P., Pearse R., et al. Perioperative cardiovascular monitoring of high-risk patients: a consensus of 12. Crit Care. 2015;19:224. doi: 10.1186/s13054-015-0932-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Bovill J.G. Intravenous anesthesia for the patient with left ventricular dysfunction. Semin Cardiothorac Vasc Anesth. 2006;10:43–48. doi: 10.1177/108925320601000108. [DOI] [PubMed] [Google Scholar]
- 29.Meng L., Yu W., Wang T., Zhang L., Heerdt P.M., Gelb A.W. Blood pressure targets in perioperative care. Hypertension. 2018;72:806–817. doi: 10.1161/HYPERTENSIONAHA.118.11688. [DOI] [PubMed] [Google Scholar]
- 30.Oneglia A., Nelson M.D., Merz C.N.B. Sex differences in cardiovascular aging and heart failure. Curr Heart Fail Rep. 2020;17:409–423. doi: 10.1007/s11897-020-00487-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Cho L., Davis M., Elgendy I., et al. Summary of updated recommendations for primary prevention of cardiovascular disease in women: JACC state-of-the-art review. J Am Coll Cardiol. 2020;75:2602–2618. doi: 10.1016/j.jacc.2020.03.060. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Mpanya D., Celik T., Klug E., Ntsinjana H. Predicting mortality and hospitalization in heart failure using machine learning: a systematic literature review. Int J Cardiol Heart Vasc. 2021;34 doi: 10.1016/j.ijcha.2021.100773. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Chicco D., Jurman G. Machine learning can predict survival of patients with heart failure from serum creatinine and ejection fraction alone. BMC Med Inform Decis Mak. 2020;20:16. doi: 10.1186/s12911-020-1023-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.McKie P.M., Kor D.J., Cook D.A., et al. Computerized advisory decision support for cardiovascular diseases in primary care: a cluster randomized trial. Am J Med. 2020;133:750–756. doi: 10.1016/j.amjmed.2019.10.039. e2. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.