Mortality following noncardiac surgery assessed by the Saint Louis University Score (SLUScore) for hypotension: a retrospective observational cohort study

Abstract

Background

The Saint Louis University Score (SLUScore) was developed to quantify intraoperative blood pressure trajectories and their associated risk for adverse outcomes. This study examines the prevalence and severity of intraoperative hypotension described by the SLUScore and its relationship with 30-day mortality in surgical subtypes.

Methods

This retrospective analysis of perioperative data included surgical cases performed between January 1, 2010, and December 31, 2020. The SLUScore is calculated from cumulative time-periods for which the mean arterial pressure is below a range of hypotensive thresholds. After calculating the SLUScore for each surgical procedure, we quantified the prevalence and severity of intraoperative hypotension for each surgical procedure and the association between intraoperative hypotension and 30-day mortality. We used binary logistic regression to quantify the potential contribution of intraoperative hypotension to mortality.

Results

We analysed 490 982 cases (57.7% female; mean age 57 yr); 33.2% of cases had a SLUScore>0, a median SLUScore of 13 (inter-quartile range [IQR] 7–21), with 1.19% average mortality. The SLUScore was associated with mortality in 12/14 surgical groups. The increases in the odds ratio for death within 30 days of surgery per SLUScore increment were: all surgery types 3.5% (95% confidence interval [95% CI] 3.2–3.9); abdominal/transplant surgery 6% (95% CI 1.5–10.7); thoracic surgery1.5% (95% CI 1–3.3); vascular surgery 3.01% (95% CI 1.9–4.05); spine/neurosurgery 1.1% (95% CI 0.1–2.1); orthopaedic surgery 1.4% (95% CI 0.7–2.2); gynaecological surgery 6.3% (95% CI 2.5–10.1); genitourinary surgery 4.84% (95% CI 3.5–6.15); gastrointestinal surgery 5.2% (95% CI 3.9–6.4); gastroendoscopy 5.5% (95% CI 4.4–6.7); general surgery 6.3% (95% CI 5.5–7.1); ear, nose, and throat surgery 1.6% (95% CI 0–3.27); and cardiac electrophysiology (including pacemaker procedures) 6.6% (95% CI 1.1–12.4).

Conclusions

The SLUScore was independently, but variably, associated with 30-day mortality after noncardiac surgery.

Editor's key points.

• •The Saint Louis University Score (SLUScore) was developed to quantify intraoperative blood pressure trajectories and their associated risk for adverse outcomes.
• •The SLUScore is calculated from cumulative periods for which the mean arterial pressure is below a range of hypotensive thresholds.
• •The authors examined the association between the SLUScore and 30-day mortality for various surgical procedures.
• •In 490 982 cases, the SLUScore was independently associated with mortality in 12/14 surgical subspecialities.
• •The association between the prevalence and severity of hypotension with 30-day postoperative mortality varied substantially across different surgical types.

Around 4.2 million people die worldwide annually within 30 days of surgery, making postoperative mortality a significant contributor to global fatalities after ischemic heart disease and stroke.¹ The rates of 30-day all-cause mortality after noncardiac surgical procedures vary with each surgery type and range from 0.5% to 3%.2, 3, 4, 5, 6, 7 Surgical procedural risk, intrinsic patient-related risk factors, and perioperative variables are associated with postoperative outcomes. Intraoperative blood pressure has been identified as an independent associate of adverse postoperative outcomes, increasing hospital length of stay and postoperative mortality.5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 The quantification of the severity of intraoperative hypotension and its association with unfavourable outcomes has been previously explored as the relationship between mean arterial pressure (MAP) and intraoperative blood pressure thresholds⁵ or as the MAP duration <65 mm Hg.²¹

Stapelfeldt and colleagues⁹ developed The Saint Louis University Score (SLUScore) based on the cumulative periods for which the effective MAP is below each of 31 hypotensive thresholds ranging from 75 to 45 mm Hg and the associated increase in the odds ratio (OR) of 30-day mortality per minute accumulated. The SLUScore allows any degree of severity, frequency, and duration of intraoperative hypotension to be quantified in relation to the associated increase in the odds of death up to 30 days after surgery. The SLUScore had been originally conceived without adjustment for the type of surgical procedure.⁹ This study elucidates the prevalence of hypotension measured by the SLUScore and its relationship with surgery-specific mortality attributed to hypotensive episodes. We hypothesised that the association of hypotension with 30-day postoperative mortality varies across different types of surgeries.

Methods

Study design

We conducted a retrospective cohort study using electronic health record data from the Indiana Network for Patient Care of noncardiac surgical cases performed between January 1, 2010, and December 31, 2020, with a known 30-day survival status. The University of Indiana Institutional Review Board (IRB # 13577) reviewed the study protocol and determined its exempt status. A data analysis and statistical plan were established before the data were accessed. Data extracted were age, sex, race, ethnicity, presence or absence of a preoperative diagnosis of hypertension (Supplementary Appendix 1), pre-procedure morbidity (Charlson Comorbidity Index [CCI]), date and duration (minutes) of surgery, type of surgery (CPT code[s]), and intraoperative MAP data. The 30-day postoperative mortality status was obtained from the Social Security Administration for 2010–2011 and from the State of Indiana Department of Health for 2012–2020.

Patient and record selection

All records of adult patients (age >18 yr) undergoing anaesthesia for noncardiac, non-obstetric surgical procedures were considered for inclusion. The following case records were excluded: (1) cases that did not allow calculation of the SLUScore; (2) successive cases that occurred within <30 days; and (3) cases with missing information on surgical duration, patient's age, CCI, procedural diagnosis, or postoperative survival status (Fig. 1).

Fig 1.

Representation of the statistical analysis performed and the study flow diagram indicating the number of patients excluded based on exclusion criteria and missing data. CCI, Charlson Comorbidity Index; SLUScore, Saint Louis University Score.

SLUScore calculation

Calculation of the SLUScore was based on the cumulative time of the MAP below each of the 31 hypotensive thresholds between 45 and 75 mm Hg. The number of minutes (exposure limits) under each of the 31 thresholds was combined into exposure limit sets that portend identical percentage increases in the OR of 30-day mortality, between 5% and 50% (Supplementary Appendix 2). The number of exposure limits exceeding the 20% limit set is the SLUScore and predicts an approximate 5% compounding progression of the odds of 30-day mortality per increment of time limit exceeded.⁹ It was determined that the exposure time limits were shorter in patients with a history of hypertension. Thus, a separate exposure limit set was created for these patients. This study calculated the SLUScore for each surgical patient offline using Microsoft Excel™ (Microsoft Corp., Redmond, WA, USA; Supplementary Appendix 2). Each MAP measurement was considered the largest of the most recent automatically acquired noninvasive MAP value (obtained within the immediately preceding 5-min period) or the most recent invasive MAP value acquired contemporaneously (continuously recorded at 1-min intervals); values of <30 mm Hg were presumed artifacts.⁹ The anaesthetic time accumulated below the 31 MAP thresholds was added for each patient and compared with the established exposure limits. The SLUScore calculation was based on different exposure limit sets in patients with a history of hypertension, as described.⁹

Surgical speciality groups

The total number of cases was grouped based on the surgical CPT codes used by the Centre for Disease Control's National Health Care Safety Network in the Surgical Site Infection Monitoring Program²² (Supplementary Appendix 1) into 13 surgical categories. Files with multiple surgical procedures and CPT codes were assigned to one category, usually the more complex one (e.g. if an orthopaedic and a vascular procedure took place in the same case, it was classified into the vascular category).

Primary outcome

The primary outcome was 30-day mortality in patients with a SLUScore >0 vs lack of hypotension (SLUScore=0) in each surgical group.

Statistical analysis

The surgical cohort and each surgical group were examined for the following data: case counts and length; patients' age, sex, race, ethnicity, and CCI; and raw 30-day mortality. Non-normally distributed data (confirmed by the Kolmogorov–Smirnov test) were expressed as median (first and third quartiles) and compared using the Kruskal–Wallis test. Normally distributed data were expressed as mean and 95% confidence interval (CI); P<0.05 was considered statistically significant. The statistical analyses were performed in SPSS (version 29.0; IBM Corp., Armonk, NY, USA). The SLUScore median was calculated for each surgical category. The incidence of hypotension (proportion SLUScore >0), the severity of hypotension (median SLUScore >0), and 30-day mortality were evaluated in the surgical cohort and all the surgical groups. The proportion of SLUScore=0 and the mortality in the SLUScore=0 and SLUScore >0 groups were calculated in the entire surgical cohort and each surgical group.

The primary analysis assessed the association between intraoperative hypotension as described by a SLUScore>0 vs lack of hypotension (SLUScore=0) and 30-day mortality with the χ² test for each surgical group. If the χ² test results were not statistically significant, a more detailed analysis was performed by exploring the association between subgroups of the SLUScore >0 groups vs SLUScore=0 and 30-day mortality. The SLUScore >0 subgroups were delimited by absolute score value: SLUScore 1–10 (named SLUScore10), SLUScore 11–20 (named SLUScore20), and SLUScore 21–31 (named SLUScore30). The association between each hypotensive subgroup vs non-hypotensive group and 30-day mortality was tested for each subgroup.

In the secondary analysis, binary logistic regression analyses were performed for each surgical speciality group to identify factors predictive of postoperative mortality. Multiplepredictive models were developed with the following variables: CCI, case duration, SLUScore, age, race, sex, and ethnicity. Only predictor variables that consistently achieved statistical significance (P<0.05) were considered in the final regression models. Adjusted OR and 95% CI for the predictor variables and concordance C-statistic and 95% CI were performed for the regression models of all surgical categories. The contribution of the SLUScore to postoperative mortality was calculated as the percent increase in the odds of 30-day mortality per SLUScore increment.

Results

Study cohort characteristics

A total of 844 673 surgical cases with at least 5-min recordings of MAP between 2010 and 2020 were identified. After applying the exclusion criteria and eliminating cases with missing information (Fig. 1), the final study cohort consisted of 490 980 surgical cases belonging to 347 106 unique patients (Table 1). The cohort's racial and ethnic distribution mirrors the demographic make-up of the state of Indiana. Variable SLUScores were calculated across specialities, with the highest median SLUScore noted in the pacemaker and EP group (9, inter-quartile range [IQR] 0–22; Fig. 2a). A total of 162 740 cases (33.15%; median age 59 yr, IQR 45–69 yr) had a SLUScore>0 (Table 2) with a SLUScore median of 13 (IQR 7–21) and median case duration of 117 min (IQR 74–184). As many as 328 240 cases (median age 56 yr, IQR 43–67 yr) had a SLUScore=0, characterised by lower mortality (0.57%) and 53-min case duration (IQR 30–100).

Table 1.

Groups of surgical procedures, case counts, case duration (minutes), patients' age, CCI median and IQR, population sex, race, and ethnicity, SLUScore and SLUScore >0 median and IQR, and mortality rate for each surgical group. AA, African American; CCI, Charlson Comorbidity Index; ENT, ear, nose, and throat; EP, electrophysiology; GI NOR, gastrointestinal non-operating room; IQR, interquartile range; SLUScore, Saint Louis University Score.

Surgical group	Number of cases	Duration 50th quartile (25th–75th quartile)	CCI 50th quartile (25th–75th quartile)	Age 50th quartile (25th–75th quartile)	Male, %	Black or AA, %	Hispanic, %	SLUS 50th quartile (25th–75th quartile)	SLUS>0 50th quartile (25th–50th quartile)	%SLUS>0, %	30-day mortality, %
All surgeries	490 890	72 (37–132)	0 (0–2)	57 (43–68)	42.30	9.99	1.94	0 (0–7)	13 (7–21)	33.15	0.78
Transplant surgery	2090	246 (209–293)	2 (0–3)	51 (40–62)	58	21	4	0 (0–10)	12 (6–19)	42.01	0.86
Ophthalmologic surgery	16 744	40 (27–60)	0 (0–1)	67 (57–75)	44.90	11.24	2.00	0 (0–0)	12 (6–18)	12.56	0.08
Thoracic surgery	3721	98 (49–186)	2 (1–4)	61 (51–70)	49	14.30	0.83	0 (0–9)	14 (6–23)	38.19	4.49
Vascular surgery	18 361	115 (63–179)	3 (1–5)	63 (53–73)	53	19.80	1.80	0 (0–15)	16 (9–25)	46.11	1.91
Spine and neurosurgery	35 486	124 (76–206)	0 (0–1)	58 (46–68)	46	6.70	1.42	0 (0–12)	14 (7–22)	44.99	1.16
Orthopaedic surgery	102 879	94 (55–132)	0 (0–1)	59 (47–69)	42	9.75	0.50	0 (0–16)	16 (9–23)	49.88	0.63
Gynaecologic surgery	44 148	63 (37–135)	0 (0–0)	41 (32–51)	0	15.40	3.10	0 (0–1)	11 (5–18)	26.76	0.07
Genitourinary surgery	52 986	69 (39–158)	1 (0–2)	62 (49–72)	67	7.71	1.76	0 (0–4)	11 (5–18)	30.52	0.60
Surgery of the GI tract	28 193	108 (60–194)	0 (0–2)	53 (40–65)	40	11.50	1.80	0 (0–5)	12 (6–21)	31.45	0.91
GI NOR endoscopy	82 233	27 (20–41)	0 (0–2)	59 (50–69)	43	5.90	1.50	0 (0–0)	8 (3–15)	9.95	0.91
General surgery	74 571	90 (55–149)	0 (0–2)	53 (40–65)	40	10.70	2.26	0 (0–8)	13 (6–20)	36.35	0.85
ENT surgery	29 028	72 (42–126)	0 (0–2)	51 (34–63)	50	10.60	2.29	0 (0–7)	13 (6–20)	34.88	0.73
Pacemaker and EP	525	177 (117–262)	1 (0–2)	59 (36–71)	65	9.33	2.80	9 (0–22)	20 (12–29)	61.52	2.86

Fig 2.

(a) Proportion of cases in the SLUScore=0 group and SLUScore >0 group for all surgeries and each surgical group. (b) Mortality rates in the SLUScore=0 and SLUScore >0 groups for all surgeries and each surgical speciality. ENT, ear, nose, and throat; EP, electrophysiology; GI NOR, gastrointestinal non-operating room; SLUScore, Saint Louis University Score; Sx, surgery.

Table 2.

Mortality rates, χ² test of association, and rates of mortality change between the SLUScore=0 and SLUScore >0 groups. ENT, ear, nose, and throat; EP, electrophysiology; GI NOR, gastrointestinal non-operating room; SLUScore, Saint Louis University Score.

Surgical group	SLUScore=0 group			SLUScore >0 group			Mortality changes between groups, %	χ2 (P-value)
	Count, n	Counts mortality, n	Mortality, %	Count, n	Counts mortality, n	Mortality, %
All surgeries	328 240	1884	0.57	162 740	1934	1.19	107.05	χ2=532.332 (<.001)
Transplant surgery	1212	6	0.50	878	12	1.37	176.08	χ2=4.53 (0.03)
Ophthalmologic surgery	14 641	11	0.08	2103	3	0.14	89.87	χ2=1.00 (0.32)
Thoracic surgery	2300	110	4.78	1421	57	4.01	–16.13	χ2=1.219 (0.27)
Vascular surgery	9895	136	1.37	8466	214	2.53	83.91	χ2=32.456 (<.001)
Spine and Neurosurgery	19 521	206	1.06	15 965	205	1.28	21.68	χ2=6.07 (0.01)
Orthopaedic surgery	51 566	250	0.48	51 313	398	0.78	59.98	χ2=34.75 (<.001)
Gynaecologic surgery	32 335	10	0.03	11 814	19	0.16	420.03	χ2=22.24 (<.001)
Genitourinary surgery	36 813	178	0.48	16 173	142	0.88	81.58	χ2=29.12 (<.001)
Surgery GI tract	19 325	107	0.55	8868	150	1.69	205.49	χ2=87.11 (<.001)
GI NOR Endoscopy	74 051	524	0.71	8182	222	2.71	283.44	χ2=329.70 (<.001)
General surgery	47 467	228	0.48	27 105	403	1.49	209.54	χ2=208.31 (<.001)
ENT surgery	18 903	116	0.61	10 125	96	0.95	54.51	χ2=10.17 (<.001)
Pacemaker and EP	202	2	0.99	323	13	4.02	306.50	χ2=4.124 (0.04)

Primary outcome: SLUScore and 30-day mortality

The SLUScore was independently associated with mortality (Fig. 2b) in all surgical specialities except for the thoracic and ophthalmologic groups (Table 2). In these two groups, a SLUScore subset analysis was performed to better understand the relationship between mortality and three subgroups of SLUScore>0: SLUScore10, SLUScore20, and SLUScore30. Only in the thoracic surgery group, the SLUScore30 subgroup showed a significant association with 30-day mortality, and hence, included for further analysis (Supplementary material).

Secondary analyses

Variables were tested for collinearity before the development of the regression models (Supplementary material). Binary logistic regression analyses were performed for each surgical speciality group where an independent association between mortality and the SLUScore was identified. The odds of 30-day mortality per SLUScore increment varied between surgical specialities (Fig. 3). The predictor variables that demonstrated a significant association with the outcome in most models were patients age, CCI, case duration, and the SLUScore. The adjusted 30-day mortality OR and 95% CI for these variables are listed in Table 3. Concordance C-statistic models with and without the SLUScore and the previously identified variables, along with the area under the receiver operating characteristic curves (ROC) and 95% CI were developed and reported for all surgical groups (Table 3).

Fig 3.

Graphical distribution of the change in the 30-day mortality OR per SLUScore increment and confidence intervals for each surgical speciality. All surgeries: OR 3.5% (95% CI 3.2–3.9), P<0.001; spine and neurosurgery: OR 1.1% (95% CI 0.1–2.1), P<0.029; orthopaedic surgery: OR 1.4 (95% CI 0.7–2.2), P<0.001; thoracic surgery: OR 1.5% (95% CI 1–3.3); thoracic surgery—SLUS30 subgroup: OR 94.1% (95% CI 24.2–203.4) P<0.004 (not represented); ENT surgery: OR 1.62% (95% CI 0–3.3), P<0.050; vascular surgery: OR 3.0% (95% CI 1.9–4.0), P<0.0001; genitourinary surgery: OR 4.8% (95% CI 3.5–6.2), P<0.001; surgery of the GI tract: OR 5.2% (95% CI 3.9–6.4), P<0.001; GI NOR endoscopy: OR 5.5% (95% CI 4.4–6.7), P<0.001; transplant surgery: OR 6% (95% CI 1.5–10.7), P<0.008; general surgery: OR 6.3% (95% CI 5.5–7.1), P<0.001; gynaecologic surgery: OR 6.3% (95% CI 2.5–10.1), P<0.001; pacemaker and EP: OR 6.6% (95% CI 1.1–12.4), P<0.017. CI, confidence interval; ENT, ear, nose, and throat; EP, electrophysiology; GI NOR, gastrointestinal non-operating room; OR, odds ratio; SLUScore, Saint Louis University Score; Sx, surgery.

Table 3.

Adjusted 30-day mortality ORs and 95% CIs for the SLUScore, age, CCI, duration, and concordance C-statistic reported with the area under the ROC and 95% CI for all surgical categories. CCI, Charlson Comorbidity Index; CI, confidence interval; ENT, ear, nose, and throat; EP, electrophysiology; GI NOR, gastrointestinal non-operating room; OR, odds ratio; ROC, receiver operating characteristic curve; SLUScore, Saint Louis University Score.

Surgical group	SLUScore, OR (95% CI)	Age, OR (95% CI)	CCI, OR (95% CI)	Duration, OR (95% CI)	ROC with SLUScore (95% CI)	ROC without SLUScore (95% CI)
All surgeries	1.035 (1.032–1.039)	1.044 (1.041–1.046)	1.24 (1.22–1.25)	1.0 (0.99–1.00)	0.796 (0.789–0.802)	0.791 (0.780–0.80)
Transplant surgery	1.06 (1.015–1.107)	1.004 (0.97–1.038)	0.938 (0.741–1.189)	1.003 (0.999–1.007)	0.648 (0.489–0.807)	0.626 (0.462–0.789)
Thoracic surgery	1.015 (0.998–1.033)	1.011 (1.00–1.023)	1.11 (1.07–1.16)	0.997 (0.995–0.999)	0.638 (0.539–0.683)	0.635 (0.590–0.680)
Thoracic SLUS30-subgroup	1.94 (1.24–3.03)
Vascular surgery	1.03 (1.019–1.04)	1.014 (1.006–1.023)	1.125 (1.086–1.166)	1.002 (1.001–1.003)	0.673 (0.643–0.702)	0.654 (0.624–0.684)
Spine and neurosurgery	1.011 (1.00–1.02)	1.029 (1.022–1.036)	1.244 (1.209–1.281)	0.999 (0.998–1.0003)	0.725(0.699–0.752)	0.726 (0.700–0.753)
Orthopaedic surgery	1.014 (1.00–1.022)	1.087 (1.08–1.094)	1.29 (1.261–1.319)	1.001 (1.000–1.002)	0.865 (0.851–0.879)	0.865 (0.851–0.879)
Gynaecologic surgery	1.063 (1.025–1.101)	1.064 (1.039–1.090)	1.313 (1.174–1.469)	1.00 (0.997–1.0043)	0.871 (0.801–0.940)	0.853 (0.773–0.932)
Genitourinary surgery	1.048 (1.035–1.061)	1.044 (1.035–1.053)	1.23 (1.194–1.266)	0.999 (0.998–1.000)	0.789 (0.766–0.813)	0.779 (0.755–0.802)
Surgery of the GI tract	1.051 (1.039–1.064)	1.058 (1.048–1.068)	1.177 (1.132–1.223)	0.998 (0.997–0.999)	0.809 (0.781–0.838)	0.796 (0.767–0.825)
GI NOR endoscopy	1.055 (1.044–1.067)	1.033 (1.027–1.039)	1.283 (1.259–1.308)	1.011 (1.0091–1.013)	0.831 (0.817–0.845)	0.823 (0.809–0.837)
General surgery	1.063 (1.055–1.071)	1.051 (1.045–1.057)	1.164 (1.135–1.194)	0.999 (0.998–1.0004)	0.807 (0.791–0.823)	0.772 (0.754–0.790)
ENT surgery	1.016 (1.00–1.032)	1.033 (1.024–1.042)	1.213 (1.168–1.259)	0.997 (0.996–0.999)	0.775 (0.747–0.803)	0.772 (0.744–0.8)
Pacemaker and EP	1.066 (1.011–1.124)	1.005 (0.974–1.036)	1.456 (1.201–1.765)	1.002 (0.998–1.006)	0.858 (0.774–0.942)	0.861 (0.791–0.931)

Discussion

We found that the prevalence, severity, and effect of hypotension on 30-day postoperative mortality varies with each type of surgery. The prevalence of hypotension (SLUScore >0) ranged from 9.95% to 61.52% across surgical groups. In the entire surgical cohort (all surgery), the 30-day mortality in the hypotensive group was double that in the non-hypotensive group (0.57% vs 1.19%).

The thoracic surgery group had the highest 30-day mortality; in this group, the 30-day mortality of the non-hypotensive and hypotensive patients was in the same range. Therefore, the impact of hypotension on mortality was challenging to evaluate and interpret. A statistically significant association was found between the SLUScore and mortality in 11 of the 13 surgical specialities studied. A SLUScore subgroup analysis revealed a strong association between the thoracic surgery group mortality and the SLUScore between 21 and 31. Besides intraoperative hypotension, multiple other factors have been previously associated with increased mortality after thoracic surgery compared with other types of surgeries (abdominal surgery), including postoperative lung injury,²³ type of surgery (pneumonectomies),^24,25 surgical pathology(lung cancer),²⁶ and institutional surgical volume.²⁶ The association between the thoracic group's 30-day mortality and only the upper bounds of the SLUScore is probably related to the high mortality in the SLUScore=0 group and other speciality-related confounders, such as disease stages and pathology (35% of patients had a diagnosis of lung cancer).

Despite the positive correlation between the SLUScore and case duration, age, and CCI, logistic regression models suggest that the SLUScore is an independent factor, raising the OR of 30-day mortality by 3.5% (95% CI 3.2–3.9) per increment in the all-surgeries group. The other statistically significant factors raising the OR of 30-day mortality and their contribution amounted to 4.4% (95% CI 4.2–4.7) per year of age and 24.2% (95% CI 23.1–25.3) per CCI unit. The 30-day mortality odds per SLUScore increment were highest in the pacemaker and EP group at 6.6% (95% CI 1.1–12.4) and the lowest in the spine and neurosurgery group at 1.1% (95% CI 0.1–2.1). The SLUScore was a predictor of mortality in 11 surgical groups. The C-statistics outcome classification was improved by including the SLUScore in the regression model in 12 of the 14 models. Six classification models performed excellently, with ROC >0.8.²⁷ Adding an outcome predictor with a significant OR may only slightly increase the ROC.²⁸ As the SLUScore was designed to identify patients at increased risk of mortality, and we had no intention of developing a prediction model, we are not interpreting the magnitude of ROC change; the higher C-scores in the SLUScore-containing models suggest that some portion of the mortality risk can be attributed to periods of hypotension. Previous studies have reported that the incidence of intraoperative hypotension increases with increasing surgery length,²⁹ patients' comorbidities (CCI scores),³⁰ patient's age,³¹ and is not less frequent in younger patients.^15,32,33 The contribution of intraoperative hypotension to postoperative 30-day mortality has been previously studied during cardiac surgery,³⁴ noncardiac surgery,^5,13,16,35 or specific surgical procedures.36, 37, 38

The reports of the incidence of intraoperative hypotension29, 30, 31^,33 vary substantially in the literature because of the lack of a uniform definition of hypotension and classification of stages of hypotension³⁹; the OR of the hypotension-associated mortality ranges from 1.8% to 20.8% or more⁴⁰ as a result of the heterogeneity of the duration and severity of hypotension events among diverse categories of surgeries.³² The strength of this retrospective analysis comes from the sample size and the method used to identify patients with hypotension who are at risk for adverse postoperative outcomes. Rather than using an arbitrary threshold, we used a previously validated score adjusted to patients' medical history of hypertension. Compared with other intraoperative hypotension assessment tools, the SLUScore shows a direct association between the severity and duration of hypotension and postoperative mortality (Supplementary material).

Our study has several important limitations, given that these data are institution and speciality-dependent, and therefore not generalisable. Although surgical procedures encompass multiple specialities, we uniquely classify them; thus, procedure misclassifications must be considered. The calculation of the SLUScore relies on a previous diagnosis of hypertension; undiagnosed hypertensive patients are at risk for classification bias. As in any retrospective analysis, errors introduced by data quality, confounding, and bias are concerns, and the extent to which they contribute is hard to assess.

In summary, the prevalence of intraoperative hypotension in this study cohort was about one-third of all cases, with hypotensive patients having double the 30-day mortality rate compared with non-hypotensive cases. The prevalence, severity, and the relationship between the magnitude of hypotension and postoperative mortality were unique to each speciality, non-linear, and influenced by factors not identified in this exploratory analysis. Although the effect size of the SLUScore contributions to postoperative mortality is small, it identifies hypotension-attributed risks that may be modified by anaesthesia providers.

Authors’ contributions

Study design; data analysis and interpretation, approval of the manuscript: CB, WHS

Drafting, reviewing, writing the first and second draft of the manuscript: CB

Manuscript revision: WHS

Declarations of interest

CB has no conflict of interest to disclose. WHS is a recipient of royalties from the Cleveland Clinic and a consultant for Edwards Lifesciences LLC.

Funding

CTSI (UL1TR002529). National Library of Medicine (2T15LM012502-06).

Handling Editor: Gareth Ackland

Appendix A. Supplementary data

The following is the Supplementary data to this article: