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 : Anesthesia & Analgesia

Secondary Logo

Journal Logo

Original Research Articles: Original Clinical Research Report

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

Hofer, Ira S. MD; Cheng, Drew MD; Grogan, Tristan MS

Author Information
doi: 10.1213/ANE.0000000000005393

Abstract

KEY POINTS

  • Question: Are key comorbid diseases accurately documented in the preanesthetic evaluation, and is a missed documentation associated with negative postoperative outcomes?
  • Findings: Key comorbid diseases are not infrequently omitted from the preanesthetic evaluation, and this missed documentation is associated with longer length of stay and postoperative mortality.
  • Meaning: There are data in the electronic health record (EHR) that are sometimes not picked up on by the anesthesiologist, and this is associated with negative postoperative outcomes.

The introduction, and subsequent proliferation, of electronic health records (EHRs) precipitated by the affordable care act and later the Health Information Technology for Economic and Clinical Health (HITECH) act has created revolutionary changes to the daily work of physicians and other care providers.1 The initial promise of the EHR was that it would lead to a consolidation of the patient’s overwhelming medical information and thereby help decrease redundant testing and incomplete information. While this consolidation has occurred, EHRs have created other issues such as distractions from interactions with patients, information overload, and physician burnout.2,3 Additionally, studies have shown that EHRs have resulted in the proliferation of redundant, and at times inaccurate information—such as copy and paste errors4—placing further challenges on providers.

Previous studies5,6 have demonstrated that despite clinicians spending a significant amount of their time going through patient’s medical records, they often miss highly important comorbid conditions. For example, 1 study found that up to 54% of patients with sepsis have never had the condition appropriately documented.7 These studies highlight the fact that while information on patient comorbidities may be present in the EHR, it is not necessarily fully apparent or reachable to the clinicians caring for these patients—the overwhelming amount of noncategorized information in the EHR makes it difficult for clinicians to quickly access relevant information. Even medications, which are often structured lists, may be incomplete or have medications prescribed for multiple overlapping conditions. For anesthesiologists, who often have a limited period of time to learn about increasingly sick patients, this is a very challenging problem. Clinically, it means that anesthesiologists may not be fully aware of conditions such as chronic kidney disease (CKD), diabetes, and other key comorbid diseases that have shown to be correlated with postoperative outcomes.8–23 Ultimately, clinicians cannot treat diseases they don’t know about, and it is possible that the failure to pick up on these comorbidities may be leading to anesthetics plans that differ in areas such as glycemic control, fluid balance, and mean arterial pressure (MAP) goals.

Previously, our group developed techniques to create rules-based algorithms that automatically phenotype patients with their comorbid diseases based on the structured data information in the patient EHR.6 We hypothesized that (1) these rules-based algorithms may detect key comorbid diseases—specifically atrial fibrillation (afib), diabetes, congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), CKD, and chronic pain—that anesthesia team did not document in the preoperative note and (2) this failure to document the diseases may be associated with worse perioperative outcomes—specifically postoperative mortality and length of stay (LOS). In this article, we first examine the incidence of missed documentation in the anesthesia preoperative evaluation by comparing the structured data from preoperative anesthesia evaluation to the results of our rules-based algorithms, and then examine the association between these missed documentation and mortality (primary outcome) and LOS (secondary outcome) by comparing patients where the algorithm and preoperative evaluation both note the disease (A+/P+) to those where the algorithm noted the disease but the preoperative evaluation did not (A+/P−).

METHODS

Data Extraction

All data for this study were extracted from the Perioperative Data Warehouse (PDW), a custom-built robust data warehouse containing all patients who have undergone surgery at the University of California Los Angeles (UCLA) Health since the implementation of our EHR (EPIC Systems, Madison, WI) in March 2013. We have previously described the creation of the PDW, which has a 2-stage design.24,25 Briefly, in the first stage, data are extracted from EPIC’s Clarity database into 29 tables organized around 3 distinct concepts: patients, surgical procedures, and health system encounters. These data are then used to populate a series of 4000 distinct measures and metrics such as procedure duration, readmissions, admission International Classification of Diseases (ICD) codes, and postoperative outcomes. All data used for this study were obtained from this data warehouse, and institutional review board approval (IRB#15-000518) was obtained from the UCLA Office of the Human Research Protection Program, including exemption from written informed consent, for this retrospective review.

Determination of Comorbidities Using Algorithms

For the purposes of this study, 6 comorbidities were selected—afib, diabetes, CHF, COPD, CKD, and chronic pain. For each disease, a rules-based algorithm designed to be highly precise (ie, few false positives) was created in line with what we have previously published.6 Briefly, in each case, an initial set of criteria for the algorithm was created. The algorithm was then run on cases in our data warehouse, and a randomly selected group of cases was chosen for review. If the algorithm was found to lack precision, then the criteria were revised until the review no longer generated false positives in our random selection. The final criteria for each disease are listed in Supplemental Digital Content 1, Table 1, https://links.lww.com/AA/D369.

Determination of Documentation of Patient Comorbidities in the Anesthesia Preoperative Evaluation

In our EHR, the preoperative anesthesia evaluation (preoperative note) is created via a series of checkboxes. Patients typically undergo one of two workflows. For approximately 60% of patients, the patients are screened by our preoperative nurses who gather the relevant documentation and prepopulate the anesthesia preoperative evaluation by clicking the various boxes and adding in the relevant text. This note is then finalized—reviewed, physical examination added, plan added, and so on—by the anesthesia team on the day of surgery. For the other 40% of patients, the note is compiled entirely by the anesthesia team on the night before and day of surgery. No distinction between these workflows was made for the purposes of this study.

For each of the comorbidities of interest, the PDW was queried to determine the checking of the appropriate box. In the event that multiple checkboxes were associated with one of the comorbidities of interest, the checking of any of the boxes was considered to be sufficient to denote anesthesiologist documentation.

Inclusion and Exclusion Criteria

Cases with anesthesia performed between April 1, 2013, (date of EHR go-live) and July 1, 2019 (date of extract creation) were included. Cases were excluded if patient age was <18 years or if the case was performed at outpatient surgery facilities (as mortality and postoperative LOS information is not relevant for these encounters).

In the event that a patient had >1 anesthetic in the database, then only the first anesthetic was selected.

Included Covariates

To independently account for associations with the outcomes of interest, a series of relevant covariates were extracted for each encounter. These included age (>89 years were set to 89 per institutional deidentification guidelines), gender, American Society of Anesthesiologists (ASA) physical status score (ASA score), emergent status of the case (booking case type), type of anesthesia (general, regional, and sedation), surgical procedure classification (Healthcare Cost and Utilization Project [HCUP] code), and preoperative score to predict postoperative mortality (POSPM) score.26

Definition of Outcome Measures

Mortality was defined as the existence of a date of death documented between a hospital admission’s admission date and discharge date. The date of death was extrapolated from documentation of a discharge disposition of “expired” and the existence of a note containing specific phrases indicative of a mortality event (“death summary,” “time of death,” “date of death,” “death date,” “death time,” and “death note”). Postoperative LOS was defined as the day of hospitalization after surgery when the patient was discharged, with a discharge on the day of surgery considered as day 1 (ie, if the patient was discharged the day after surgery it would be day 2, etc).

Statistical Analysis

Patient characteristics and study outcomes were summarized overall and between the comorbidity classification groups using frequency (%) or mean (standard deviation [SD]). For mortality, all logistic regression models were constructed with terms for POSPM score, ASA, booking case type, anesthesia type, HCUP code, and the 4-level diagnosis variable we created (ie, algorithm positive/preoperative evaluation positive [A+/P+], algorithm positive/preoperative evaluation negative [A+/P−], etc). Similarly, linear models were constructed with the same terms for our other outcome of interest, postoperative LOS. Since the distributional assumptions of these models were questionable, we are reporting the models after log transformation of the outcome (LOS) and report the results as geometric mean ratios (the ratios of the antilogs of the mean values of the log-transformed data). On further residual analysis, the normality assumption seemed adequate, but we did observe possible homoscedasticity violations, so we decided to use robust standard errors (SAS option empirical = mbn in proc glimmix). For the models including diabetes, we also included A1c and home insulin usage and for chronic pain models, we included preoperative opioid usage and pain clinic visit (yes/no). Pairwise comparisons (or odds ratios/geometric mean ratios) between the comorbidity classification groups (A+/P+, A+/P−, etc) were extracted from the models and presented with 95% confidence intervals (CIs). Forest plots were constructed to aid visual interpretation of the results.

We also ran models for each outcome using the total number of missed diagnoses to get an aggregate overall result (see Tables 1 and 2/Figure). Analyses were conducted using SAS V9.4 (SAS Institute, Cary, NC). As we ran 12 models (6 comorbidities × 2 outcomes), our overall α level used for determining significance was a Bonferroni-adjusted .004 (.05/12 = .004).

F1
Figure.:
Plots showing the effects and 95% confidence interval of not documented comorbid diseases on postoperative mortality (A) (odds ratio) and LOS (B) (ratio of days increased). The incremental effect of each undocumented disease is shown with the red box at the bottom. For mortality, afib was associated with an increased risk; each additional undocumented disease increased the odds of mortality. For LOS, all diseases except chronic pain were associated with an increased LOS (geometric mean ratio >1), and each additional undocumented disease was also associated with an increased LOS. afib indicates atrial fibrillation; CI, confidence interval; LOS indicates length of stay.

An a priori sample size calculation was not performed as we extracted all records that met inclusion criteria.

RESULTS

Table 1. - Key Demographic Information
CHF Diabetes afib COPD CKD Chronic pain
Variable A+/P+ A−/P+ A+/P− A−/P− A+/P+ A−/P+ A+/P− A−/P− A+/P+ A−/P+ A+/P− A−/P− A+/P+ A−/P+ A+/P− A−/P− A+/P+ A−/P+ A+/P− A−/P− A+/P+ A−/P+ A+/P− A−/P− Total
Patients 3296 (3.6%) 599 (0.7%) 3929 (4.3%) 83,187 (91.4%) 11,894 (13.1%) 1890 (2.1%) 3625 (4.0%) 73,602 (80.9%) 2340 (2.6%) 231 (0.3%) 7028 (7.7%) 81,412 (89.5%) 2800 (3.1%) 1005 (1.1%) 3934 (4.3%) 83,272 (91.5%) 2228 (2.4%) 1565 (1.7%) 9847 (10.8%) 77,371 (85.0%) 2170 (2.4%) 1715 (1.9%) 31,439 (34.5%) 55,687 (61.2%) 91,011
ASA physical status
 I 0 (0%) 0 (0%) 5 (0.1%) 6866 (8.3%) 14 (0.1%) 12 (0.6%) 20 (0.6%) 6825 (9.3%) 1 (0%) 0 (0%) 10 (0.1%) 6860 (8.4%) 1 (0%) 2 (0.2%) 27 (0.7%) 6841 (8.2%) 0 (0%) 7 (0.4%) 44 (0.4%) 6820 (8.8%) 20 (0.9%) 56 (3.3%) 1742 (5.5%) 5053 (9.1%) 6871 (7.5%)
 II 72 (2.2%) 31 (5.2%) 177 (4.5%) 35,990 (43.3%) 2116 (17.8%) 563 (29.8%) 609 (16.8%) 32,982 (44.8%) 281 (12%) 36 (15.6%) 790 (11.2%) 35,163 (43.2%) 268 (9.6%) 203 (20.2%) 787 (20%) 35,012 (42%) 230 (10.3%) 308 (19.7%) 1268 (12.9%) 34,464 (44.5%) 688 (31.7%) 733 (42.7%) 10,549 (33.6%) 24,300 (43.6%) 36,270 (39.9%)
 III 1845 (56%) 452 (75.5%) 2291 (58.3%) 35,805 (43%) 8001 (67.3%) 1183 (62.6%) 2179 (60.1%) 29,030 (39.4%) 1614 (69%) 176 (76.2%) 4196 (59.7%) 34,407 (42.3%) 1998 (71.4%) 713 (70.9%) 2146 (54.6%) 35,536 (42.7%) 1476 (66.2%) 1043 (66.6%) 6467 (65.7%) 31,407 (40.6%) 1354 (62.4%) 877 (51.1%) 15,450 (49.1%) 22,712 (40.8%) 40,393 (44.4%)
 IV 1340 (40.7%) 112 (18.7%) 1377 (35%) 4148 (5%) 1726 (14.5%) 122 (6.5%) 775 (21.4%) 4354 (5.9%) 436 (18.6%) 18 (7.8%) 1952 (27.8%) 4571 (5.6%) 528 (18.9%) 85 (8.5%) 947 (24.1%) 5417 (6.5%) 508 (22.8%) 207 (13.2%) 1891 (19.2%) 4371 (5.6%) 108 (5%) 49 (2.9%) 3563 (11.3%) 3257 (5.8%) 6977 (7.7%)
 V 39 (1.2%) 3 (0.5%) 72 (1.8%) 349 (0.4%) 37 (0.3%) 10 (0.5%) 40 (1.1%) 376 (0.5%) 8 (0.3%) 1 (0.4%) 77 (1.1%) 377 (0.5%) 5 (0.2%) 2 (0.2%) 27 (0.7%) 429 (0.5%) 13 (0.6%) 0 (0%) 156 (1.6%) 294 (0.4%) 0 (0%) 0 (0%) 131 (0.4%) 332 (0.6%) 463 (0.5%)
 VI 0 (0%) 1 (0.2%) 7 (0.2%) 29 (0%) 0 (0%) 0 (0%) 2 (0.1%) 35 (0%) 0 (0%) 0 (0%) 3 (0%) 34 (0%) 0 (0%) 0 (0%) 0 (0%) 37 (0%) 1 (0%) 0 (0%) 21 (0.2%) 15 (0%) 0 (0%) 0 (0%) 4 (0%) 33 (0.1%) 37 (0%)
Booking class
 Blank/missing 281 (8.5%) 68 (11.4%) 365 (9.3%) 6808 (8.2%) 811 (6.8%) 208 (11%) 318 (8.8%) 6185 (8.4%) 228 (9.7%) 31 (13.4%) 635 (9%) 6628 (8.1%) 200 (7.1%) 66 (6.6%) 327 (8.3%) 6929 (8.3%) 167 (7.5%) 90 (5.8%) 792 (8%) 6473 (8.4%) 150 (6.9%) 105 (6.1%) 2498 (7.9%) 4769 (8.6%) 7522 (8.3%)
 Critically emergent 19 (0.6%) 2 (0.3%) 38 (1%) 959 (1.2%) 54 (0.5%) 19 (1%) 47 (1.3%) 898 (1.2%) 10 (0.4%) 1 (0.4%) 66 (0.9%) 941 (1.2%) 8 (0.3%) 6 (0.6%) 22 (0.6%) 982 (1.2%) 8 (0.4%) 6 (0.4%) 155 (1.6%) 849 (1.1%) 2 (0.1%) 0 (0%) 217 (0.7%) 799 (1.4%) 1018 (1.1%)
 Elective 1707 (51.8%) 433 (72.3%) 1870 (47.6%) 59,492 (71.5%) 7804 (65.6%) 1381 (73.1%) 1864 (51.4%) 52,453 (71.3%) 1554 (66.4%) 176 (76.2%) 4080 (58.1%) 57,692 (70.9%) 1885 (67.3%) 820 (81.6%) 2275 (57.8%) 58,522 (70.3%) 1252 (56.2%) 1209 (77.3%) 4429 (45%) 56,612 (73.2%) 1737 (80%) 1542 (89.9%) 19,033 (60.5%) 41,190 (74%) 63,502 (69.8%)
 Emergent 146 (4.4%) 25 (4.2%) 267 (6.8%) 4275 (5.1%) 412 (3.5%) 119 (6.3%) 194 (5.4%) 3988 (5.4%) 62 (2.6%) 15 (6.5%) 368 (5.2%) 4268 (5.2%) 109 (3.9%) 35 (3.5%) 173 (4.4%) 4396 (5.3%) 127 (5.7%) 33 (2.1%) 616 (6.3%) 3937 (5.1%) 48 (2.2%) 15 (0.9%) 1589 (5.1%) 3061 (5.5%) 4713 (5.2%)
 Inpatient 618 (18.8%) 18 (3%) 691 (17.6%) 3543 (4.3%) 1053 (8.9%) 21 (1.1%) 470 (13%) 3326 (4.5%) 243 (10.4%) 3 (1.3%) 846 (12%) 3778 (4.6%) 266 (9.5%) 26 (2.6%) 480 (12.2%) 4098 (4.9%) 303 (13.6%) 107 (6.8%) 898 (9.1%) 3562 (4.6%) 119 (5.5%) 19 (1.1%) 3345 (10.6%) 1387 (2.5%) 4870 (5.4%)
 Transplant case 100 (3%) 32 (5.3%) 139 (3.5%) 2286 (2.7%) 672 (5.6%) 86 (4.6%) 291 (8%) 1508 (2%) 28 (1.2%) 1 (0.4%) 255 (3.6%) 2273 (2.8%) 66 (2.4%) 16 (1.6%) 228 (5.8%) 2247 (2.7%) 107 (4.8%) 24 (1.5%) 1863 (18.9%) 563 (0.7%) 18 (0.8%) 19 (1.1%) 1025 (3.3%) 1495 (2.7%) 2557 (2.8%)
 Urgent 425 (12.9%) 21 (3.5%) 559 (14.2%) 5824 (7%) 1088 (9.1%) 56 (3%) 441 (12.2%) 5244 (7.1%) 215 (9.2%) 4 (1.7%) 778 (11.1%) 5832 (7.2%) 266 (9.5%) 36 (3.6%) 429 (10.9%) 6098 (7.3%) 264 (11.8%) 96 (6.1%) 1094 (11.1%) 5375 (6.9%) 96 (4.4%) 15 (0.9%) 3732 (11.9%) 2986 (5.4%) 6829 (7.5%)
Sex
 Female 1323 (40.1%) 222 (37.1%) 1685 (42.9%) 44,124 (53%) 5286 (44.4%) 900 (47.6%) 1634 (45.1%) 39,534 (53.7%) 831 (35.5%) 94 (40.7%) 2647 (37.7%) 43,782 (53.8%) 1208 (43.1%) 434 (43.2%) 2028 (51.6%) 43,684 (52.5%) 865 (38.8%) 561 (35.8%) 4652 (47.2%) 41,276 (53.3%) 1259 (58%) 976 (56.9%) 16,804 (53.4%) 28,315 (50.8%) 47,354 (52%)
 Male 1973 (59.9%) 377 (62.9%) 2244 (57.1%) 39,063 (47%) 6608 (55.6%) 990 (52.4%) 1991 (54.9%) 34,068 (46.3%) 1509 (64.5%) 137 (59.3%) 4381 (62.3%) 37,630 (46.2%) 1592 (56.9%) 571 (56.8%) 1906 (48.4%) 39,588 (47.5%) 1363 (61.2%) 1004 (64.2%) 5195 (52.8%) 36,095 (46.7%) 911 (42%) 739 (43.1%) 14,635 (46.6%) 27,372 (49.2%) 43,657 (48%)
Anesthesia type
 Mac/regional/other 1303 (39.5%) 199 (33.2%) 1290 (32.8%) 17,996 (21.6%) 3028 (25.5%) 506 (26.8%) 1000 (27.6%) 16,254 (22.1%) 841 (35.9%) 69 (29.9%) 2119 (30.2%) 17,759 (21.8%) 777 (27.8%) 262 (26.1%) 1110 (28.2%) 18,639 (22.4%) 644 (28.9%) 371 (23.7%) 2514 (25.5%) 17,259 (22.3%) 413 (19%) 239 (13.9%) 8184 (26%) 11,952 (21.5%) 20,788 (22.8%)
 General 1993 (60.5%) 400 (66.8%) 2639 (67.2%) 65,191 (78.4%) 8866 (74.5%) 1384 (73.2%) 2625 (72.4%) 57,348 (77.9%) 1499 (64.1%) 162 (70.1%) 4909 (69.8%) 63,653 (78.2%) 2023 (72.3%) 743 (73.9%) 2824 (71.8%) 64,633 (77.6%) 1584 (71.1%) 1194 (76.3%) 7333 (74.5%) 60,112 (77.7%) 1757 (81%) 1476 (86.1%) 23,255 (74%) 43,735 (78.5%) 70,223 (77.2%)
Surgical service
 Cardiac surgery 739 (22.4%) 50 (8.3%) 992 (25.2%) 2248 (2.7%) 963 (8.1%) 27 (1.4%) 442 (12.2%) 2597 (3.5%) 228 (9.7%) 3 (1.3%) 1476 (21%) 2322 (2.9%) 236 (8.4%) 30 (3%) 515 (13.1%) 3248 (3.9%) 219 (9.8%) 98 (6.3%) 773 (7.9%) 2939 (3.8%) 31 (1.4%) 23 (1.3%) 2067 (6.6%) 1908 (3.4%) 4029 (4.4%)
 Cardiology 905 (27.5%) 118 (19.7%) 847 (21.6%) 3498 (4.2%) 797 (6.7%) 68 (3.6%) 282 (7.8%) 4221 (5.7%) 747 (31.9%) 3 (1.3%) 1908 (27.1%) 2710 (3.3%) 232 (8.3%) 38 (3.8%) 329 (8.4%) 4769 (5.7%) 226 (10.1%) 99 (6.3%) 897 (9.1%) 4146 (5.4%) 45 (2.1%) 26 (1.5%) 1518 (4.8%) 3779 (6.8%) 5368 (5.9%)
 Gastroenterology 324 (9.8%) 59 (9.8%) 382 (9.7%) 8542 (10.3%) 1323 (11.1%) 357 (18.9%) 481 (13.3%) 7146 (9.7%) 203 (8.7%) 28 (12.1%) 506 (7.2%) 8570 (10.5%) 306 (10.9%) 151 (15%) 575 (14.6%) 8275 (9.9%) 227 (10.2%) 159 (10.2%) 909 (9.2%) 8012 (10.4%) 161 (7.4%) 114 (6.6%) 3361 (10.7%) 5671 (10.2%) 9307 (10.2%)
 General surgery 185 (5.6%) 37 (6.2%) 283 (7.2%) 12,943 (15.6%) 1623 (13.6%) 239 (12.6%) 418 (11.5%) 11,168 (15.2%) 184 (7.9%) 28 (12.1%) 512 (7.3%) 12,724 (15.6%) 291 (10.4%) 116 (11.5%) 422 (10.7%) 12,619 (15.2%) 269 (12.1%) 221 (14.1%) 890 (9%) 12,068 (15.6%) 249 (11.5%) 230 (13.4%) 4154 (13.2%) 8815 (15.8%) 13,448 (14.8%)
 Neurosurgery 3 (0.1%) 0 (0%) 5 (0.1%) 41 (0%) 10 (0.1%) 0 (0%) 1 (0%) 38 (0.1%) 2 (0.1%) 0 (0%) 2 (0%) 45 (0.1%) 1 (0%) 0 (0%) 2 (0.1%) 46 (0.1%) 1 (0%) 1 (0.1%) 6 (0.1%) 41 (0.1%) 1 (0%) 0 (0%) 24 (0.1%) 24 (0%) 49 (0.1%)
 Obstetrics and gynecology 44 (1.3%) 15 (2.5%) 34 (0.9%) 6590 (7.9%) 375 (3.2%) 77 (4.1%) 88 (2.4%) 6143 (8.3%) 38 (1.6%) 1 (0.4%) 61 (0.9%) 6583 (8.1%) 47 (1.7%) 17 (1.7%) 102 (2.6%) 6517 (7.8%) 37 (1.7%) 23 (1.5%) 219 (2.2%) 6404 (8.3%) 102 (4.7%) 62 (3.6%) 2808 (8.9%) 3711 (6.7%) 6683 (7.3%)
 Orthopedics 231 (7%) 30 (5%) 306 (7.8%) 10,484 (12.6%) 1245 (10.5%) 116 (6.1%) 390 (10.8%) 9300 (12.6%) 236 (10.1%) 17 (7.4%) 492 (7%) 10,306 (12.7%) 296 (10.6%) 71 (7.1%) 377 (9.6%) 10,307 (12.4%) 238 (10.7%) 179 (11.4%) 813 (8.3%) 9821 (12.7%) 446 (20.6%) 213 (12.4%) 4703 (15%) 5689 (10.2%) 11,051 (12.1%)
 Others 263 (8%) 89 (14.9%) 502 (12.8%) 15,533 (18.7%) 2047 (17.2%) 368 (19.5%) 705 (19.4%) 13,267 (18%) 220 (9.4%) 53 (22.9%) 816 (11.6%) 15,298 (18.8%) 446 (15.9%) 153 (15.2%) 674 (17.1%) 15,114 (18.2%) 301 (13.5%) 258 (16.5%) 1457 (14.8%) 14,371 (18.6%) 519 (23.9%) 350 (20.4%) 5734 (18.2%) 9784 (17.6%) 16,387 (18%)
 Otolaryngology 141 (4.3%) 56 (9.3%) 121 (3.1%) 7495 (9%) 1019 (8.6%) 251 (13.3%) 172 (4.7%) 6371 (8.7%) 135 (5.8%) 34 (14.7%) 305 (4.3%) 7339 (9%) 337 (12%) 186 (18.5%) 321 (8.2%) 6969 (8.4%) 131 (5.9%) 232 (14.8%) 386 (3.9%) 7064 (9.1%) 246 (11.3%) 332 (19.4%) 1539 (4.9%) 5696 (10.2%) 7813 (8.6%)
 Plastic surgery 14 (0.4%) 8 (1.3%) 16 (0.4%) 1922 (2.3%) 101 (0.8%) 22 (1.2%) 25 (0.7%) 1812 (2.5%) 13 (0.6%) 0 (0%) 28 (0.4%) 1919 (2.4%) 13 (0.5%) 9 (0.9%) 33 (0.8%) 1905 (2.3%) 15 (0.7%) 14 (0.9%) 62 (0.6%) 1869 (2.4%) 33 (1.5%) 30 (1.7%) 605 (1.9%) 1292 (2.3%) 1960 (2.2%)
 Surgical oncology 24 (0.7%) 12 (2%) 33 (0.8%) 1567 (1.9%) 195 (1.6%) 60 (3.2%) 41 (1.1%) 1340 (1.8%) 34 (1.5%) 8 (3.5%) 55 (0.8%) 1539 (1.9%) 43 (1.5%) 18 (1.8%) 32 (0.8%) 1543 (1.9%) 41 (1.8%) 27 (1.7%) 125 (1.3%) 1443 (1.9%) 44 (2%) 71 (4.1%) 485 (1.5%) 1036 (1.9%) 1636 (1.8%)
 Thoracic surgery 21 (0.6%) 5 (0.8%) 49 (1.2%) 1578 (1.9%) 195 (1.6%) 14 (0.7%) 50 (1.4%) 1394 (1.9%) 30 (1.3%) 1 (0.4%) 245 (3.5%) 1377 (1.7%) 134 (4.8%) 42 (4.2%) 201 (5.1%) 1276 (1.5%) 32 (1.4%) 21 (1.3%) 134 (1.4%) 1466 (1.9%) 42 (1.9%) 17 (1%) 780 (2.5%) 814 (1.5%) 1653 (1.8%)
 Urology 177 (5.4%) 65 (10.9%) 193 (4.9%) 8450 (10.2%) 1367 (11.5%) 235 (12.4%) 372 (10.3%) 6911 (9.4%) 164 (7%) 39 (16.9%) 371 (5.3%) 8311 (10.2%) 236 (8.4%) 122 (12.1%) 206 (5.2%) 8321 (10%) 353 (15.8%) 177 (11.3%) 2383 (24.2%) 5972 (7.7%) 173 (8%) 133 (7.8%) 2689 (8.6%) 5890 (10.6%) 8885 (9.8%)
 Vascular surgery 225 (6.8%) 55 (9.2%) 166 (4.2%) 2296 (2.8%) 634 (5.3%) 56 (3%) 158 (4.4%) 1894 (2.6%) 106 (4.5%) 16 (6.9%) 251 (3.6%) 2369 (2.9%) 182 (6.5%) 52 (5.2%) 145 (3.7%) 2363 (2.8%) 138 (6.2%) 56 (3.6%) 793 (8.1%) 1755 (2.3%) 78 (3.6%) 114 (6.6%) 972 (3.1%) 1578 (2.8%) 2742 (3%)
Postoperative LOS, median (Q1–Q3) 4 (2–9) 2 (1–5) 5 (2–10) 2 (1–4) 3 (2–7) 2 (1–4) 4 (2–8) 2 (1–4) 2 (1–5) 2 (1–3) 4 (2–10) 2 (1–5) 3 (1–7) 2 (1–4) 3 (2–8) 2 (1–5) 4 (2–8) 2 (1–5) 5 (2–9) 2 (1–4) 3 (1–5) 2 (1–4) 3 (1–6) 2 (1–4) 2 (1–5)
Mortality 143 (4.3%) 11 (1.8%) 220 (5.6%) 718 (0.9%) 207 (1.7%) 25 (1.3%) 125 (3.4%) 735 (1%) 41 (1.8%) 6 (2.6%) 311 (4.4%) 734 (0.9%) 70 (2.5%) 15 (1.5%) 124 (3.2%) 883 (1.1%) 87 (3.9%) 33 (2.1%) 415 (4.2%) 557 (0.7%) 18 (0.8%) 5 (0.3%) 555 (1.8%) 514 (0.9%) 1092 (1.2%)
Abbreviations: afib, atrial fibrillation; ASA, American Society of Anesthesiologists; CHF, congestive heart failure; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; LOS, length of stay.

Overall, the data set included 91,011 cases. Age ranged from 18 to 89 years. There were 52% women and 48% men. A total of 70% of the cases involved patients admitted from home and 2.8% of cases were transplants. The cases involved a variety of surgical services with 10.2% gastrointestinal procedures, 14.8% general surgical procedures, and 12.1% orthopedic procedures. Around 7.5% of cases had ASA physical status I, 39.9% of cases ASA II, 44.4% ASA III, 7.7% ASA IV, and 0.5% ASA V. Complete details of patient demographics can be found in Table 1.

Missed Documentation Incidence

Overall the agreement between the algorithms and the preoperative note was >84% for all comorbidities other than chronic pain (63.5%). However, when there were divergent results, most often it was the algorithm-detecting disease that was missed in the preoperative note. The algorithm-detected disease not documented by the anesthesia team in 34.5% of cases for chronic pain (vs 1.9% of cases where chronic pain was documented but not detected by the algorithm), 4.0% of cases for diabetes (vs 2.1%), 4.3% of cases for CHF (vs 0.7%), 4.3% of cases for COPD (vs 1.1%), 7.7% of cases for afib (vs 0.3%), and 10.8% of cases for CKD (vs 1.7%).

Association of Undocumented Disease With LOS and Mortality

To assess if the undocumented disease itself was associated with worse postoperative outcomes, we created risk-adjusted models for both LOS and mortality. The results of the risk-adjusted models for mortality and LOS were compared for cases where the preprocedure evaluation did not note the disease but the algorithm did flag the disease (P−/A+) with cases where the algorithm and the preoperative note both flagged the disease (P+/A+). Thus, in both groups, the disease was present in the patient, the difference was the documentation in the preoperative evaluation. This analysis demonstrated that for all diseases except chronic pain, the lack of documentation of the disease on the preoperative evaluation was associated with a longer LOS. With regard to mortality, the discrepancy was associated with increased mortality for afib with the remainder of the P values lacking significance. These results are shown in Table 2 and the Figure. The comparisons for all groups are shown in Supplemental Digital Content 2, Table 2, https://links.lww.com/AA/D370.

Table 2. - Adjusted Risk of Mortality and AKI for Each Missed Diagnosis
Mortality LOS (log scale)
Missed diagnosis Odds ratio (95% CI) P value Geo mean ratio (95% CI) P value
Afib 1.75 (1.23-2.48) .002 1.25 (1.20-1.28) <.001
Chronic pain 0.95 (0.58-1.56) .846 0.98 (0.95-1.02) .339
Diabetes 1.27 (0.98-1.63) .066 1.06 (1.03-1.10) <.001
Congestive heart failure 1.21 (0.96-1.54) .114 1.11 (1.06-1.15) <.001
Chronic obstructive pulmonary disease 1.13 (0.82-1.56) .456 1.11 (1.06-1.15) <.001
Chronic kidney disease 0.98 (0.75-1.27) .860 1.06 (1.02-1.10) .002
Each additional missed diagnosis 1.52 (1.42-1.63) <.001 1.11 (1.10-1.12) <.001
We expect a 25% increase in the geometric mean of LOS for afib missed diagnosis compared to A+/P+. For each additional missed diagnosis, the geometric mean of LOS increases about 11%.
Abbreviations: afib, atrial fibrillation; CI, confidence interval; LOS, length of stay.

Models were also created to examine any additive effect of missing multiple diagnoses. For each missed disease, the odds of mortality increased 1.52 (95% CI, 1.42-1.63) and the LOS increased by approximately 11%, with a geometric mean ratio of 1.11 (95% CI, 1.10-1.12). These results are shown in the Figure and Table 2.

DISCUSSION

In this article, we used precise criteria in rules-based algorithms for common comorbid diseases and examined cases where the algorithm-identified disease that was not documented in the anesthesia preprocedure evaluation. Overall, the concordance between the algorithms and the preoperative note was good for all diseases other than chronic pain (Table 1). However, in those cases where the algorithm-detected disease that was not documented in the note, there was an association with an increased LOS for all diseases and an increased mortality for afib. Additionally, the effect was additive for both mortality and LOS—that is, each additional disease that was not documented was associated with a longer LOS and mortality (Figure and Table 2).

The findings of this study bring some important take-away points. First, while the EHR has helped solve the problem of physician’s lack of access to key information about their patients, the information is often spread across multiple locations, poorly presented, and at times hard to determine. Thus, while the information may be technically available, it is often infeasible for physicians to access given their time constraints and the realities of modern patient care. Consistent with previous studies,6,7 we show that key diseases are not documented with some frequency.

Of particular note is the discrepancy with regard to chronic pain. A total of 34.5% of patients were flagged as having chronic pain by the algorithms, yet this condition was not documented in the preoperative note. The reasons for this are unclear. One possibility is that the criteria used in the algorithm were too sensitive—for example, occasional home opioid use may not always indicate chronic pain. However, all of the criteria (home opioid use, previous documentation of chronic pain syndromes, previous visits to pain clinics, and previous consultation by pain management) are consistent with risk factors for increased postoperative pain. Thus, another possibility is that chronic pain is particularly underrecognized by anesthesia providers. This could potentially be due to the lack of a clear definition as opposed to many other more “objective” diseases. Given the current opioid epidemic and recurring issues around postoperative pain, this may warrant future study.

The second point is that this failure to document the existence of key comorbid diseases is associated with worse perioperative outcomes—specifically mortality and LOS. It has been well reported that diseases such as diabetes, CHF, COPD, afib, and others are associated with worse perioperative outcomes. What has been less proven is the extent to which these risk factors might be modifiable. While some studies looking at better glucose control,27 or prehabilitation28 have shown positive effects on postoperative outcomes, most studies have been limited to single diseases or small in size. Furthermore, the objective management of some of these diseases may not necessarily differ from a regular patient—that is, a patient with well-controlled CHF or paroxysmal afib may not have anything specific optimized perioperatively. Nonetheless, we have noted that in patients where the condition was not documented in the preoperative evaluation, there was an association with increased LOS and the number of diseases missed was associated with longer LOS and mortality. A growing body of evidence suggests that what we as anesthesiologists do in the operating room has profound effects on longer term outcomes. For example, recent evidence has demonstrated intraoperative hypotension to be associated with postoperative acute kidney injury (AKI) and mortality,13,22 blood transfusion with postoperative pulmonary and thromboembolic complications and infection,29 glucose management with postoperative infections,27 and lung-protective ventilation with postoperative pulmonary complications.30 Furthermore, enhanced recovery after surgery (ERAS) protocols, which are often a distillation of best practices, have been demonstrated to improve postoperative outcomes.31 An exhaustive list of best practice management for each of the comorbidities that were studied and the extent to which they were applied to each patient would be beyond the scope of this article. We also did not attempt to differentiate whether patients, where the anesthesiologist did not document the disease, had any different postoperative treatment by the surgical team. Nonetheless, this study does demonstrate that the difficulties associated with simply obtaining accurate information from the EHR are associated with worse perioperative outcomes.

There are some important limitations of this study. This is a single-center retrospective review. Thus, it is possible that these findings may not extend to other locations. However, our EHR (EPIC) is the most widely used EHR in the United States. While some elements of EHR design are customized by the hospital, the overall issues of usability are widespread and well documented. Thus, we believe that while the exact incidence of missing disease documentation might differ, most hospitals will have similar issues.

Another limitation of this study is the way in which we determined the documentation of the key comorbid diseases in the preoperative evaluation. As noted in the methods, in our EHR, the preoperative note is populated by a series of checkboxes for common and significant diseases. However, the note also contains areas for free text. Thus, it is possible that in some patients, the disease was documented in free text and not with a checkbox, thus overestimating the incidence of missed documentation. Similarly, it is possible the anesthesia team was aware of the disease, but their documentation was incomplete—something that cannot be quantified in a retrospective chart review. For this reason, throughout the article, we have chosen to use the term “missed documentation” as opposed to missed diagnosis. There is likely a reasonable degree of correlation between the comorbidities the anesthesia team is aware of and those documented—the degree of correlation is beyond the scope of this study. However, as clinicians, we do believe that there is at least some fraction of patients for whom a comorbid disease is unknown by the anesthesia team at the time of the procedure. While this limitation would result in an overestimation of the incidence of missed disease, it would actually serve to reduce the power of detecting an association between the missed disease and the outcomes of interest—LOS and mortality. The fact that these associations were present, despite this limitation, supports the hypothesis that the associations between a missed comorbidity and perioperative outcomes are real and that the choices anesthesiologists make in managing these critical diseases have implications beyond the operating room.

Finally, this retrospective study cannot demonstrate causation, only association. While we attempted to adjust for all clinical confounders, we cannot rule out the existence of other confounding variables, bias, and so on.

Overall, these results add to the increasing body of evidence that current workflows create challenges for physicians to accurately and quickly obtain and document critical information about comorbid patient diseases. Additionally, the association between this missed documentation and postoperative outcomes augment the many studies that indicate that the anesthetic plan has effects well beyond the operating room in ways that we have not yet fully understand. This study supports the trend toward increased utilization of care pathways and user interface design to improve the EHR and enhance patient safety.

DISCLOSURES

Name: Ira S. Hofer, MD.

Contribution: This author helped with study design, statistical analysis, and manuscript preparation.

Conflicts of Interest: I. S. Hofer is the president of Clarity Healthcare Analytics Inc, a company that assists hospitals with extracting and using data from their electronic medical records. The company currently owns the rights to the PDW software that was used to extract data from the electronic health record. I. S. Hofer receives research funding from Merck Pharmaceuticals.

Name: Drew Cheng, MD.

Contribution: This author helped extract the data.

Conflicts of Interest: None.

Name: Tristan Grogan, MS.

Contribution: This author helped with statistical analysis.

Conflicts of Interest: None.

This manuscript was handled by: Thomas M. Hemmerling, MSc, MD, DEAA.

    REFERENCES

    1. Tai-Seale M, Olson CW, Li J, et al. Electronic health record logs indicate that physicians split time evenly between seeing patients and desktop medicine. Health Aff (Millwood). 2017;36:655–662.
    2. Shanafelt TD, Dyrbye LN, Sinsky C, et al. Relationship between clerical burden and characteristics of the electronic environment with physician burnout and professional satisfaction. Mayo Clin Proc. 2016;91:836–848.
    3. Downing NL, Bates DW, Longhurst CA. Physician burnout in the electronic health record era: are we ignoring the real cause? Ann Intern Med. 2018;169:50–51.
    4. Weis JM, Levy PC. Copy, paste, and cloned notes in electronic health records: prevalence, benefits, risks, and best practice recommendations. Chest. 2014;145:632–638.
    5. Gabel E, Hofer IS, Satou N, et al. Creation and validation of an automated algorithm to determine postoperative ventilator requirements after cardiac surgery. Anesth Analg. 2017;124:1423–1430.
    6. Hofer IS, Cheng D, Grogan T, et al. Automated assessment of existing patient’s revised cardiac risk index using algorithmic software. Anesth Analg. 2019;128:909–916.
    7. Jolley RJ, Quan H, Jetté N, et al. Validation and optimisation of an ICD-10-coded case definition for sepsis using administrative health data. BMJ Open. 2015;5:e009487.
    8. Hsieh JK, Dalton JE, Yang D, Farag ES, Sessler DI, Kurz AM. The association between mild intraoperative hypotension and stroke in general surgery patients. Anesth Analg. 2016;123:933–939.
    9. van Waes JA, Grobben RB, Nathoe HM, et al.; Cardiac Health After Surgery (CHASE) Investigators. One-year mortality, causes of death, and cardiac interventions in patients with postoperative myocardial injury. Anesth Analg. 2016;123:29–37.
    10. Khan J, Alonso-Coello P, Devereaux PJ. Myocardial injury after noncardiac surgery. Curr Opin Cardiol. 2014;29:307–311.
    11. Devereaux PJ, Biccard BM, Sigamani A, et al.; Writing Committee for the VISION Study Investigators. Association of postoperative high-sensitivity troponin levels with myocardial injury and 30-day mortality among patients undergoing noncardiac surgery. JAMA. 2017;317:1642–1651.
    12. Botto F, Alonso-Coello P, Chan MT, et al.; Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Writing Group, on behalf of The Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Investigators; Appendix 1. The Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Study Investigators Writing Group; Appendix 2. The Vascular events In noncardiac Surgery patIents cOhort evaluatioN Operations Committee; Vascular events In noncardiac Surgery patIents cOhort evaluatioN VISION Study Investigators. Myocardial injury after noncardiac surgery: a large, international, prospective cohort study establishing diagnostic criteria, characteristics, predictors, and 30-day outcomes. Anesthesiology. 2014;120:564–578.
    13. Salmasi V, Maheshwari K, Yang D, et al. Relationship between intraoperative hypotension, defined by either reduction from baseline or absolute thresholds, and acute kidney and myocardial injury after noncardiac surgery: a retrospective cohort analysis. Anesthesiology. 2017;126:47–65.
    14. Walsh M, Garg AX, Devereaux PJ, Argalious M, Honar H, Sessler DI. The association between perioperative hemoglobin and acute kidney injury in patients having noncardiac surgery. Anesth Analg. 2013;117:924–931.
    15. Oprea AD, Lombard FW, Liu WW, et al. Baseline pulse pressure, acute kidney injury, and mortality after noncardiac surgery. Anesth Analg. 2016;123:1480–1489.
    16. Walsh M, Devereaux PJ, Garg AX, et al. Relationship between intraoperative mean arterial pressure and clinical outcomes after noncardiac surgery: toward an empirical definition of hypotension. Anesthesiology. 2013;119:507–515.
    17. Sun LY, Wijeysundera DN, Tait GA, Beattie WS. Association of intraoperative hypotension with acute kidney injury after elective noncardiac surgery. Anesthesiology. 2015;123:515–523.
    18. Brienza N, Giglio MT, Marucci M, Fiore T. Does perioperative hemodynamic optimization protect renal function in surgical patients? A meta-analytic study. Crit Care Med. 2009;37:2079–2090.
    19. Monk TG, Saini V, Weldon BC, Sigl JC. Anesthetic management and one-year mortality after noncardiac surgery. Anesth Analg. 2005;100:4–10.
    20. Stapelfeldt WH, Yuan H, Dryden JK, et al. The SLUScore: a novel method for detecting hazardous hypotension in adult patients undergoing noncardiac surgical procedures. Anesth Analg. 2017;124:1135–1152.
    21. Bijker JB, van Klei WA, Vergouwe Y, et al. Intraoperative hypotension and 1-year mortality after noncardiac surgery. Anesthesiology. 2009;111:1217–1226.
    22. Monk TG, Bronsert MR, Henderson WG, et al. Association between intraoperative hypotension and hypertension and 30-day postoperative mortality in noncardiac surgery. Anesthesiology. 2015;123:307–319.
    23. Mascha EJ, Yang D, Weiss S, Sessler DI. Intraoperative mean arterial pressure variability and 30-day mortality in patients having noncardiac surgery. Anesthesiology. 2015;123:79–91.
    24. Epstein RH, Hofer IS, Salari V, Gabel E. Successful implementation of a perioperative data warehouse using another hospital’s published specification from Epic’s electronic health record system. Anesth Analg. 2021;132:465–474.
    25. Hofer IS, Gabel E, Pfeffer M, Mahbouba M, Mahajan A. A systematic approach to creation of a perioperative data warehouse. Anesth Analg. 2016;122:1880–1884.
    26. Le Manach Y, Collins G, Rodseth R, et al. Preoperative Score to Predict Postoperative Mortality (POSPOM): derivation and validation. Anesthesiology. 2016;124:570–579.
    27. Ehrenfeld JM, Wanderer JP, Terekhov M, Rothman BS, Sandberg WS. A perioperative systems design to improve intraoperative glucose monitoring is associated with a reduction in surgical site infections in a diabetic patient population. Anesthesiology. 2017;126:431–440.
    28. Carli F, Feldman LS. From preoperative risk assessment and prediction to risk attenuation: a case for prehabilitation. Br J Anaesth. 2019;122:11–13.
    29. Glance LG, Dick AW, Mukamel DB, et al. Association between intraoperative blood transfusion and mortality and morbidity in patients undergoing noncardiac surgery. Anesthesiology. 2011;114:283–292.
    30. Futier E, Constantin JM, Paugam-Burtz C, et al. A trial of intraoperative low-tidal-volume ventilation in abdominal surgery. N Engl J Med. 2013;369:428–437.
    31. Group EC. The impact of enhanced recovery protocol compliance on elective colorectal cancer resection: results from an international registry. Ann Surg. 2015;261:1153–1159.

    Supplemental Digital Content

    Copyright © 2021 International Anesthesia Research Society