In-hospital coded prevalence and associated factors of postoperative delirium following rotator cuff repair surgery: a retrospective national inpatient sample database study

Abstract

Background

While postoperative delirium (POD) is a well-known postoperative complication, research on its incidence and risk factors in rotator cuff repair surgeries lacks a robust analysis using a comprehensive national dataset.

Methods

Leveraging the National Inpatient Sample (NIS), the most extensive all-payer hospital database in the U.S., we performed a retrospective cohort study analyzing 7,216 patients who underwent rotator cuff repair from 2016 to 2019. Our assessment included patient demographics, hospital-related variables, comorbid conditions, and postoperative adverse events to identify key associations.

Results

Among patients undergoing inpatient rotator cuff repair between 2016 and 2019, the in-hospital coded prevalence of POD was 7.5%, with the highest rate (8.8%) recorded in 2019. Patients who developed POD following rotator cuff repair surgery demonstrated several adverse outcomes, including advanced age, increased comorbidity burden, prolonged hospital stays, and higher hospitalization costs (p < 0.001). They were also more prone to experiencing various perioperative medical and surgical complications during their hospitalization. These complications encompassed urinary tract infections, pneumonia, acute renal failure, septicemia, thrombocytopenia, blood transfusions, acute cerebrovascular events, peripheral vascular disease, urinary retention, and other medical concerns. Surgical complications, such as hemorrhage, seroma, or hematoma, were also observed. Multiple factors were associated with POD, including advanced age (65 years or older), Medicaid or private insurance coverage, alcohol or drug abuse, deficiency anemias, neurological disorders, and psychoses. On the other hand, being female and having an elective hospital admission reduced the risk of developing POD.

Conclusion

Among patients undergoing inpatient rotator cuff repair, POD had an in-hospital coded prevalence of 7.5%—a non-negligible proportion. Studying its contributing factors remains important, and recognizing these risks may help clinicians refine treatment approaches and enhance patient recovery. These findings are not generalizable to outpatient rotator cuff repair procedures.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13741-026-00709-x.

Background

In the United States, shoulder-related conditions place a considerable strain on healthcare resources (Luime et al. 2004). By 2000, the annual direct medical costs linked to these disorders had already reached a staggering $7 billion (Karjalainen et al. 2019). Among these conditions, rotator cuff pathologies are the leading contributor, accounting for 65% to 85% of cases, though prevalence varies by patient demographics and clinical context (Chard et al. 1991; Ostör et al. 2005; Vecchio et al. 1995). As a result, surgical treatments have gained traction among medical professionals (Karjalainen et al. 2019). Research from the UK highlights a dramatic surge in rotator cuff repairs and subacromial decompression surgeries, with age-adjusted rates jumping from 1.4 to 13.7 per 100,000 people between 2004 and 2010 (Gomoll et al. 2004). Despite the increasing prevalence of these surgical interventions, a notable proportion of postoperative patients are encountering complications.

Postoperative delirium (POD) has been established as a common complication after major orthopedic procedures like rotator cuff repair (Rhee et al. 2025). This neurological condition manifests as diminished cognitive function, altered consciousness or perception, attentional deficits, disorientation, and memory deficits (Rai et al. 2014; Maclullich et al. 2008; Inouye et al. 2015). The onset of postoperative delirium has been attributed to the brain's dysfunctional adaptation to surgical stress (Maclullich et al. 2008; Urbánek et al. 2023), imposing substantial burdens on patients and healthcare systems (Leslie et al. 2008). Associated consequences include elevated mortality rates, progressive functional decline, long-term cognitive disorders, and other serious complications (Scott et al. 2015; Jin et al. 2020; Goldberg et al. 2020). Moreover, delirium significantly escalates medical expenditures and prolongs hospitalization (Jin et al. 2020; Goldberg et al. 2020; Gleason et al. 2015). In the US, total annual direct healthcare costs linked to delirium among elderly patients are estimated at $143 billion to $152 billion (Leslie et al. 2008; Inouye et al. 2016). Beyond acute POD episodes, perioperative cognitive disturbances may also emerge as postoperative cognitive decline (POCD) following discharge (Daiello et al. 2019). Distinct from delirium, POCD is not a formal clinical diagnosis; instead, it reflects a measurable reduction in cognitive test performance relative to preoperative baselines (Feinkohl et al. 2023). Notably, POD serves as a strong predictor for subsequent POCD development, suggesting that delirium prevention strategies could potentially mitigate long-term cognitive impairment risk (Glumac et al. 2019).

Consequently, the preoperative identification of patients at high risk of POD is essential for improving postoperative outcomes and preventing complications (Kalisvaart et al. 2006; Nandi et al. 2014). Current evidence suggests that advanced age, significant surgical risks, and the necessity for comprehensive pharmacologic pain management may drive POD development in orthopedic surgery (Papaioannou et al. 2005; Bruce et al. 2007). Studies have also identified several risk factors associated with postoperative delirium, such as Parkinson’s disease, psychiatric conditions, cognitive decline, depression, diabetes, and imbalances in fluids or electrolytes after surgery (Oh et al. 2015; Wang et al. 2016; de Jong et al. 2019). However, no large-scale national database analysis has hitherto explored the incidence or risk factors of POD specifically following rotator cuff repair procedures. Utilizing a comprehensive national database in the United States, this research investigated the occurrence and associated risk factors of postoperative delirium in patients undergoing rotator cuff repair. Although POD incidence is expected to be relatively low in this surgical population, numerous contributing factors may influence its development. Overall, our study highlights the importance of identifying high-risk patients to inform targeted preoperative interventions.

Methods

Data source

This study utilized the National Inpatient Sample (NIS), the largest all-payer US inpatient database, which forms part of the Healthcare Cost and Utilization Project (HCUP). Sponsored by the Agency for Healthcare Research and Quality (AHRQ), HCUP represents the most extensive collection of American hospital care data. The NIS employs a stratified sampling design, encompassing over 1,000 hospitals across 46 states; this sample constitutes approximately 20% of annual US hospital admissions (Liu et al. 2024). Our analysis extracted the following variables for each sampled hospitalization: patient demographics, admission status, clinical outcomes (length of stay [LOS]), along with diagnoses, procedures, and comorbidities coded using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM).

Data collection

The study population was identified from the NIS database, which included all adult patients (aged 18 years or older) who underwent rotator cuff repair surgery between January 1, 2016, and December 31, 2019. Notably, a substantial number of rotator cuff repairs are performed in outpatient or ambulatory surgery settings and are therefore not represented in this cohort. Patients were selected based on the presence of the relevant ICD-10-CM procedure codes for rotator cuff repair (Fig. 1). According to the ICD-10-CM diagnostic code (supplement table), patients with delirium were diagnosed and selected including transient, acute, and subacute delirium (F05, F06.0, F06.2, R41.841-R41.843, R41.89), drug-induced delirium (F11.921, F11.951, F13.921), altered mental state (R41.82) (Zheng et al. 2025).

Fig. 1.

Flow diagram of study selection process. ICD-10, International Classification of Diseases (Tenth Revision) Clinical Modification

The patients were divided into two cohorts based on the presence or absence of POD. Demographic characteristics, such as age, race, and sex, as well as outcome measures including admission status, length of stay, total hospitalization costs, insurance type, and 25 comorbidity variables (as captured in the NIS database), were compared between the two groups (Table 1).

Table 1.

Variables used in binary logistic regression analysis

Variables Categories	Specific Variables
Patient demographics	Age (≤ 64 years and ≥ 65 years), race (White, Black, Hispanic, Asian or Pacific Islander, Native American and Other)
Hospital characteristics	Type of insurance (Medicare, Medicaid, private insurance, self-pay, no charge, other), bed size of hospital (small, medium, large), teaching status of hospital (nonteaching, teaching), type of admission (non-elective, elective), location of the hospital (northeast, Midwest or north central, south, west)
Comorbidities	Smoking, AIDS, alcohol abuse, deficiency anemias, rheumatoid arthritis, chronic blood loss anemia, congestive heart failure, chronic pulmonary disease, coagulopathy, depression, diabetes, drug abuse, hypertension, hypothyroidism, liver disease, lymphoma, fluid and electrolyte disorders, metastatic cancer, other neurological disorders, obesity, paralysis, peripheral vascular disorders, psychoses, pulmonary circulation disorders, renal failure
Complications	Nausea and vomitting, blood transfusion, urinary tract infection, acute renal failure, thrombocytopenia, genitourinary disease, pneumonia, DVT, peripheral vascular disease, septicemia, acute cerebrovascular disease, stroke, urinary retention, hemorrhagic anemia, upper limb nerve injury, hemorrhage/seroma/hematoma, wound infection, disruption of wound

The ICD-10-CM diagnostic codes were used to identify perioperative medical and surgical complications that may be independently associated with the development of POD prior to hospital discharge. The medical complications included nausea and vomiting, blood transfusion, urinary tract infection, acute renal failure, thrombocytopenia, genitourinary disorders, pneumonia, deep vein thrombosis, peripheral vascular disease, septicemia, acute cerebrovascular events, stroke, urinary retention, and hemorrhagic anemia. The surgical complications comprised upper limb nerve injury, hemorrhage/seroma/hematoma, wound infection, and wound disruption.

Data analysis

All statistical analyses were performed using SPSS version 25 software. To examine the distribution of quantitative variables, including age, length of stay, and total hospitalization costs, the Kolmogorov–Smirnov test for normality was conducted. For data with a non-parametric distribution, non-parametric tests were conducted. Continuous variables were compared between groups using the Wilcoxon rank-sum test, whereas categorical variables were assessed via the chi-square test. To explore the association between POD and perioperative complications, both univariate and multivariate logistic regression models were constructed, applying the Enter method. Each model was constructed independently, without mutual adjustment across models, to avoid overadjustment and reverse causality. Model 1 (baseline patient characteristics) included patient demographics and hospital characteristics (Table 1). Model 2 (comorbidities) included smoking, AIDS, alcohol abuse, deficiency anemias, etc. (Table 1). Model 3 (postoperative complications) included nausea and vomitting, blood transfusion, urinary tract infection, etc. (Table 1). Because the NIS does not provide the timing of complications relative to delirium onset, Model 3 was interpreted separately and these variables are referred to as “concurrent in-hospital events” rather than risk factors. For all tests, statistical significance was set at p < 0.05.

Results

In-hospital coded prevalence of postoperative delirium in patients undergoing rotator cuff repair surgery among 7,216 rotator cuff repair procedures identified in the NIS database (2016–2019), 539 patients experienced POD, yielding an overall prevalence of 7.5% (Table 2). Furthermore, a rising trend in POD in-hospital coded prevalence was noted over this period, increasing from 6.10% in 2016 to 8.80% in 2019 (Fig. 2).

Table 2.

Patient characteristics and outcomes after rotator cuff repair surgery (2015–2019)

Characteristics	POD	No POD	p
Total (n = count)	539	6677
Total incidence (%)	7.5
Age (median, years)	66 (56, 73)	62 (54, 71)	< 0.001
Age group (%)
18–44	15.0	9.7	< 0.001
45–64	35.1	38.2
65–74	23.0	31.8
≥ 75	26.9	20.3
Gender (%)
Female	46.4	49.5	0.160
Male	53.6	50.5
Race (%)
White	79.8	78.9	0.548
Black	9.3	8.6
Hispanic	7.2	8.1
Asian or Pacific Islander	1.5	1.1
Native American	0.6	0.5
Other	1.7	2.8
Number of Comorbidity (%)
0	4.6	12.0	< 0.001
1	11.7	20.1
2	16.1	21.7
≥ 3	67.5	46.3
TOTCHG (median, $)	72,729 (45,902.5–139,594)	63,115 (42,262–99,414)	< 0.001
Type of insurance (%)
Medicare	60.3	54.4	< 0.001
Medicaid	16.0	9.2
Private insurance	18.2	26.9
Self-pay	2.0	2.3
No charge	0.2	0.3
Other	3.3	6.9
Region of hospital (%)
Northeast	16.3	17.0	0.953
Midwest or North Central	21.7	21.2
South	39.9	39.1
West	22.1	22.7
Bed size of hospital (%)
Small	22.6	23.3	0.937
Medium	28.2	27.9
Large	49.2	48.7
Type of hospital (teaching %)	73.8	70.2	0.078
Elective admission (%)	43.0	67.1	< 0.001
LOS (median, d)	4 (2–9)	2 (1–4)	< 0.001

P0D Postoperative delirium, LOS Length of stay, TOTCHE Total charge

Fig. 2.

Annual incidence of POD after rotator cuff repair surgery

Demographic and hospital characteristics

A significant age disparity was observed between cohorts, with POD patients demonstrating a median age of 66 years compared to 62 years in non-POD cases (p < 0.001). Furthermore, the distribution varied significantly between the groups, with a notably higher POD prevalence in individuals over 75 (p < 0.001) (Table 3). Comorbidity burden also differed, as patients with a Charlson Comorbidity Index (CCI) ≥ 3 had a substantially higher POD in-hospital coded prevalence (67.5% vs. 46.3%). Delirium patients experienced longer hospital stays (median 4 days vs. 2 days) and incurred higher total hospitalization costs (median $72,729 vs. $63,115), reflecting an increase of $9,614. Moreover, the distribution of insurance types differed between groups: the proportion of patients with Medicare was higher among those with POD (60.3%) compared to those without POD (54.4%). The likelihood of POD patients undergoing elective admission was relatively low (43.0% vs. 67.1%). In addition, there were no significant differences between gender, race, region of hospital, bed size of hospital, and type of hospital (Table 2).

Table 3.

Risk factors associated with POD after rotator cuff repair surgery

Variable	Multivariate Logistic Regression
	OR	95% CI	p
Age ≥ 65 years old	1.785	1.463—2.178	< 0.001
Female	0.791	0.645—0.97	0.024
Race
White	Ref	——	——
Black	1.641	0.799—3.373	0.177
Hispanic	1.459	0.669—3.181	0.342
Asian or Pacific Islander	1.258	0.570—2.776	0.570
Native American	2.150	0.749—6.177	0.155
Other	1.015	0.239—4.302	0.984
Number of Comorbidity
0	Ref	——	——
1	1.481	0.906—2.420	0.117
2	1.837	1.106—3.052	0.019
≥ 3	2.678	1.510—4.751	0.001
Type of insurance
Medicare	Ref	——	——
Medicaid	2.228	1.315—3.775	0.003
Private insurance	2.088	1.187—3.673	0.011
Self-pay	1.402	0.820—2.395	0.217
No charge	1.026	0.446—2.357	0.952
Other	0.616	0.071—5.381	0.662
Bed size of hospital
Small	Ref	——	——
Medium	0.864	0.864—1.126	0.280
Large	0.786	0.786—1.003	0.053
Elective admission	0.583	0.466—0.730	< 0.001
Teaching hospital	1.080	0.870—1.341	0.485
Region of hospital
Northeast	Ref	——	——
Midwest or North Central	0.954	0.699—1.303	0.768
South	1.107	0.838—1.462	0.473
West	0.951	0.692—1.306	0.756

POD Postoperative delirium, OR Odds ratio, CI Confidence interval

Relationship between other postoperative complications and POD

Patients undergoing rotator cuff repair surgery who experienced POD exhibited a higher prevalence of blood transfusion (4.3%), urinary tract infection (2.7%), acute renal failure (6.2%), septicemia (5.1%), acute cerebrovascular disease (0.6%), DVT (0.7%), thrombocytopenia (2.1%), genitourinary disease (6.9%), pneumonia (2.5%), peripheral vascular disease (3.1%), urinary retention (2.9%), stroke (0.6%), and hemorrhagic anemia (11.4%) than those without POD (p ≤ 0.001) (Table 4 and Fig. 3). Besides, multiple regression analysis suggested that POD was related to urinary tract infection (OR = 1.986; CI = 1.345–2931) and urinary retention (OR = 2.275; CI = 1.546–3.348) (Table 4 and Fig. 3).

Table 4.

Relationship between POD and rotator cuff repair surgery

Complications	Univariate Analysis			Multivariate Logistic Regression
	POD	No POD	p	OR	95% CI	p
Medical complications
Nausea and Vomiting	4 (0.7%)	117 (1.8%)	0.079	0.408	0.145–1.145	0.088
Blood transfusion	51 (9.5%)	286 (4.3%)	< 0.001	1.222	0.845–1.767	0.287
Urinary tract infection	48 (8.9%)	181 (2.7%)	< 0.001	1.986	1.345–2.931	0.001
Acute renal failure	79 (14.7%)	413 (6.2%)	< 0.001	1.831	0.587–5.716	0.298
Thrombocytopenia	20 (3.7%)	143 (2.1%)	0.018	0.517	0.241–1.111	0.091
Genitourinary disease	84 (15.6%)	458 (6.9%)	< 0.001	0.564	0.185–1.722	0.315
Pneumonia	29 (5.4%)	164 (2.5%)	< 0.001	1.304	0.800–2.126	0.287
DVT	12 (2.2%)	46 (0.7%)	< 0.001	1.575	0.741–3.347	0.237
Peripheral vascular disease	30 (5.6%)	206 (3.1%)	0.002	1.156	0.534–2.503	0.713
Septicemia	65 (12.1%)	338 (5.1%)	< 0.001	0.796	0.545–1.162	0.237
Acute cerebrovascular disease	12 (2.2%)	41 (0.6%)	< 0.001	3.795	0.484–29.738	0.204
Urinary retention	40 (7.4%)	192 (2.9%)	< 0.001	2.275	1.546–3.348	< 0.001
Stroke	12 (2.2%)	41 (0.6%)	< 0.001	0.540	0.067–4.342	0.563
Hemorrhagic anemia	100 (18.6%)	763 (11.4%)	< 0.001	1.241	0.950–1.622	0.113
Surgical complications
Upper limb nerve injury	2 (0.4%)	15 (0.2%)	0.832	1.547	0.331–7.217	0.579
Hemorrhage/seroma/hematoma	8 (1.5%)	64 (1.0%)	< 0.001	0.948	0.412–2.177	0.899
Wound infection	9 (1.7%)	102 (1.5%)	0.796	1.083	0.508–2.310	0.837
Disruption of wound	7 (1.3%)	73 (1.1%)	0.661	0.793	0.335–1.876	0.598

POD Postoperative delirium, DVT deep vein thrombosis, OR Odds ratio, CI Confidence interval

Fig. 3.

Postoperative complications associated with POD after rotator cuff repair surgery

In-hospital events associated with POD after rotator cuff repair surgery

To identify in-hospital events associated with POD, a logistic regression analysis was conducted, revealing several significant predictors (Tables 3 and 5). Key associated factors included advanced age (≥ 65 years, OR = 1.785, 95% CI = 1.463–2.178), Medicaid coverage (OR = 2.228, CI = 1.315–3.775), private insurance (OR = 2.088, CI = 1.187–3.673), alcohol abuse (OR = 2.446, CI = 1.709–3.503), deficiency anemias (OR = 1.584, CI = 1.016–2.471), drug abuse (OR = 2.326, CI = 1.610–3.360), other neurological disorders (OR = 4.480, CI = 3.249–6.178), and psychoses (OR = 2.343, CI = 1.596–3.439). Conversely, two protective factors were identified: female sex (OR = 0.791, CI = 0.645–0.97, p < 0.001) and elective admission (OR = 0.583, CI = 0.466–0.730, p < 0.001) (Figs. 4 and 5).

Table 5.

Relationship between POD and preoperative comorbidities

Comorbidities	Univariate Analysis			Multivariate Logistic Regression
	POD	No POD	p	OR	95% CI	p
Preoperative comorbidities
Smoking	179 (36.9%)	2198 (32.9%)	0.058	0.969	0.789–1.190	0.764
AIDS	2 (0.4%)	31 (0.5%)	0.758	0.453	0.101–2.035	0.302
Alcohol abuse	57 (10.6%)	219 (3.3%)	< 0.001	2.446	1.709–3.503	< 0.001
Deficiency anemias	32 (5.9%)	141 (2.1%)	< 0.001	1.584	1.016–2.471	0.043
Rheumatoid arthritis	22 (4.1%)	287 (4.3%)	0.811	0.904	0.566–1.444	0.672
Chronic blood loss anemia	5 (0.9%)	31 (0.5%)	0.250	1.095	0.378–3.170	0.867
Congestive heart failure	65 (12.1%)	466 (7.0%)	< 0.001	1.392	0.921–2.104	0.116
Chronic pulmonary disease	112 (20.8%)	1248 (18.7%)	0.233	1.019	0.799–1.301	0.877
Coagulopathy	39 (7.2%)	230 (3.4%)	< 0.001	1.682	0.934–3.027	0.083
Depression	97 (18.0%)	952 (14.3%)	0.018	1.214	0.945–1.559	0.130
Diabetes	74 (13.7%)	649 (9.7%)	0.003	1.026	0.757–1.390	0.870
Drug abuse	65 (12.1%)	264 (4.0%)	< 0.001	2.326	1.610–3.360	< 0.001
Hypertension	343 (63.6%)	4079 (61.1%)	0.243	1.068	0.861–1.325	0.547
Hypothyroidism	86 (16.0%)	931 (13.9%)	0.197	1.164	0.893–1.517	0.261
Liver disease	48 (8.9%)	298 (4.5%)	< 0.001	0.949	0.643–1.402	0.794
Lymphoma	5 (0.9%)	29 (0.4%)	0.200	1.947	0.679–5.584	0.215
Fluid and electrolyte disorders	149 (27.6%)	895 (13.4%)	< 0.001	1.159	0.897–1.499	0.260
Metastatic cancer	8 (1.5%)	109 (1.6%)	0.793	0.694	0.293–1.642	0.406
Other neurological disorders	85 (15.8%)	176 (2.6%)	< 0.001	4.480	3.249–6.178	< 0.001
Obesity	91 (16.9%)	1349 (20.2%)	0.640	0.826	0.636–1.073	0.152
Paralysis	20 (3.7%)	102 (1.5%)	< 0.001	1.525	0.871–2.670	0.140
Peripheral vascular disorders	29 (5.4%)	207 (3.1%)	0.004	1.054	0.482–2.303	0.895
Psychoses	44 (8.2%)	181 (2.7%)	< 0.001	2.343	1.596–3.439	< 0.001
Pulmonary circulation disorders	9 (1.7%)	86 (1.3%)	0.454	0.756	0.352–1.623	0.472
Renal failure	67 (12.4%)	492 (7.4%)	< 0.001	1.257	0.902–1.751	0.178

POD Postoperative delirium, AIDS Acquired Immune Deficiency Syndrome, OR Odds ratio, CI Confidence interval

Fig. 4.

Risk factors associated with POD after rotator cuff repair surgery

Fig. 5.

Preoperative comorbidities associated with POD after rotator cuff repair surgery

Discussion

This study presents findings from a comprehensive health economics analysis focusing on POD in rotator cuff repair patients. Between 2016 and 2019, POD in-hospital coded prevalence increased steadily from 6.10% to 8.80% (Fig. 2), a trend not previously documented. Two key factors may explain this rise: first, evolving diagnostic criteria (ICD-10-CM) and clinician awareness, as POD identification can vary with expertise despite consistent clinical definitions. Second, the growing elderly population has led to an increase in rotator cuff surgeries (1976; Peng et al. 2024). However, factors such as limited surgical understanding, suboptimal medical interventions, and evolving anesthesia or pain management protocols may contribute to higher POD rates (Fineberg et al. 2013; Peng et al. 2024; Bozic et al. 2013).

Interestingly, in this study, the in-hospital coded prevalence of POD after rotator cuff repair was 7.5%. For context, reported rates of POD following major orthopedic surgeries such as total joint arthroplasty or hip fracture repair typically range from 5 to 17% (Xiao et al. 2023; Igwe et al. 2023). Given that rotator cuff repair is generally associated with lower surgical stress and shorter operative time, the observed 7.5% is unexpectedly high for this specific procedure. This finding likely reflects the inherent limitations of our inpatient-only cohort (which selects for higher-risk patients) and potential overcapture or miscoding of delirium using ICD-10 codes (Igwe et al. 2023; Sun and Yang 2023).

Demographic analysis revealed that the median age of POD patients was 4 years higher than that of non-POD cases (Table 2). Logistic regression further identified advanced age (≥ 65 years) as an independent predictor of POD risk (Table 3). This aligned with clinical observations of POD's prevalence in elderly populations and established literature recognizing age as a key predictor (Mevorach et al. 2023; Sadeghirad et al. 2023). The underlying mechanism may involve age-related physiological vulnerabilities: brain atrophy, reduced neuronal density, diminished cerebral blood flow, and altered neurotransmitter levels collectively increase delirium susceptibility (Turner et al. 2016; Yang et al. 2020). Notably, emerging evidence implicates gut microbiota in neurological pathogenesis. Perioperative interventions may induce murine POD-like behaviors through age-dependent microbiota shifts, including Lactobacilli depletion. Consequently, postoperative microbiota alterations in elderly patients represent a potential contributor to neurocognitive dysfunction (Peng et al. 2024; Liufu et al. 2020; Zhang et al. 2023).

The elevated comorbidity burden among POD patients reflects their poorer preoperative health status and heightened postoperative complication risk (Mevorach et al. 2023; Sadeghirad et al. 2023). Our analysis corroborates established findings linking POD to increased financial costs and hospitalization duration (Table 2).

Specifically, POD extended median hospital stays by 4 days, incurring an additional US$9,614 per admission. However, we must acknowledge a potential bidirectional relationship. While it is plausible that POD directly contributes to extended hospitalization and increased resource use (e.g., due to falls, aspiration, or delayed rehabilitation), the opposite direction is equally possible: patients with longer admissions or greater baseline illness severity are more likely to undergo delirium screening and have POD documented in discharge codes. This detection bias or reverse causality could partly explain the observed associations (Fineberg et al. 2013; Peng et al. 2024). Therefore, our estimates should be viewed as reflecting associations rather than purely causal effects of POD on LOS and costs. Future prospective studies with standardized delirium assessments at predefined time points are needed to disentangle these relationships.

Logistic regression identified elective admission as a protective factor against POD (Fig. 4), which may reflect better baseline health or more comprehensive preoperative optimization in scheduled admissions, thereby reducing vulnerability to postoperative complications, including POD (Yang et al. 2021). Notably, female gender demonstrated protective effects (Fig. 4), though the underlying mechanisms remain unclear. Previous research suggests potential involvement of pineal hormone melatonin, which exhibits metabolic links with estrogen (Campbell et al. 2019; de Jonghe et al. 2011). Besides, altered melatonin metabolism or circadian rhythm disruptions may contribute to POD pathogenesis (Campbell et al. 2019; de Jonghe et al. 2011).

Although some risk factors for POD, like older age and pre-existing neurological or psychiatric conditions, are non-modifiable, healthcare providers can still educate patients about these risks and closely monitor those more vulnerable. On the other hand, contributing factors such as alcohol misuse, anemia, and substance abuse may be partially addressed through interventions before surgery to help reduce the likelihood of POD development. Optimizing these potentially modifiable comorbidities can enhance a patient's readiness for the elective rotator cuff repair surgery (Aldecoa et al. 2024).

One noteworthy but difficult-to-interpret finding is the apparent association between insurance type (Medicaid and private insurance) and higher odds of POD. Given the observational nature of the NIS and the absence of key clinical and socioeconomic covariates, we caution readers against attributing causality to this association. Age, admission type, hospital characteristics, regional coding variations, and unmeasured social determinants likely account for much of the observed effect. As such, these results should be viewed as exploratory and warrant validation in datasets with more granular patient-level information.

Several important constraints must be acknowledged in this investigation, with particular emphasis on the substantial limitations for causal inference imposed by the NIS database. First, the NIS exclusively captures data documented prior to patient discharge, which may result in underestimation of POD events and cannot capture post-discharge complications. Moreover, because the database does not provide the precise timing of delirium documentation relative to surgery, our estimates reflect the in-hospital coded prevalence of POD rather than a true incidence rate; the term “incidence” would be misleading in this context. Second, and most critically, the NIS lacks information on a wide range of established associated factors and confounders for POD, including baseline dementia, frailty, functional dependence, anesthetic technique (e.g., intravenous vs. inhalation), use of regional blocks, perioperative sedatives (e.g., benzodiazepines), cumulative opioid exposure, and the specific delirium screening method employed. The absence of these variables fundamentally prevents us from adjusting for significant confounders, thereby severely restricting any claims of causality. Consequently, our findings can only be interpreted as demonstrating associations rather than causal effects, and the observed estimates may be substantially biased due to unmeasured confounding (Oh et al. 2015; Mevorach et al. 2023; Liu et al. 2024). Third, like other large administrative databases, the NIS is susceptible to coding inconsistencies and documentation inaccuracies. Administrative datasets typically demonstrate high specificity but low sensitivity for detecting adverse outcomes such as POD, which could lead to non-differential misclassification and potentially bias our results toward the null (Fineberg et al. 2013; Bozic et al. 2013). Fourth, and importantly for generalizability, the study population is restricted to inpatient admissions captured by the NIS. A substantial and increasing proportion of rotator cuff repair surgeries are performed in outpatient or ambulatory surgery settings, which are not represented in this database. Therefore, our cohort consists of a selected, higher-risk subgroup—likely including patients with greater comorbidity burdens or more complex surgical indications. This selection bias means that our estimated POD prevalence of 7.5% is not generalizable to the broader population of rotator cuff repair patients and likely overestimates the true POD burden. Readers should exercise caution when extrapolating our findings to all rotator cuff repair procedures. Fifth, temporal ambiguity precludes causal interpretation of associations with postoperative complications. The NIS database records diagnoses based on the entire hospital stay but does not provide the chronological order of events. Therefore, for variables such as postoperative pneumonia, acute kidney injury, or respiratory failure, we cannot ascertain whether these complications occurred before, concurrent with, or after the onset of POD. Consequently, we have refrained from labeling these as “risk factors” and instead describe them as “associated in-hospital events.” Readers should not infer causality from these associations. Furthermore, the associations between POD and longer hospital stays or higher costs may be subject to a bidirectional relationship. While POD may contribute to prolonged hospitalization and increased resource use, longer admissions or greater illness severity also increase the likelihood of POD being screened, documented, and coded. This reverse causality or detection bias precludes any firm causal inference regarding the economic impact of POD in this administrative dataset. Given these inherent limitations, our results should be viewed as hypothesis-generating. Future prospective studies incorporating outpatient populations, detailed clinical assessments, and standardized delirium screening protocols are urgently needed.

Conclusions

POD following rotator cuff repair surgery is a costly complication, affecting 7.5% of patients, with a steady rise in annual in-hospital coded prevalence between 2016 and 2019. Associated factors identified in this study include non-elective admissions, alcohol or drug abuse, deficiency anemias, neurological conditions, and psychoses, along with associated complications like urinary tract infections and retention. Interestingly, advanced age (≥ 65 years) and male sex are associated with increased risk, while younger female patients and elective admissions were associated with a lower risk of POD. However, the reasons underlying these protective factors warrant further investigation. Furthermore, POD is correlated with prolonged hospital stays and increased healthcare expenses.

Abbreviation

POD

Postoperative Delirium

ICD-10-CM

International Classification of Diseases, Tenth Revision, Clinical Modification

NIS

National Inpatient Sample

HCUP

Healthcare Cost and Utilization Project

AHRQ

Agency for Healthcare Research and Quality

ICU

Intensive Care Unit

LOS

Length of Stay

OR

Odds ratio

CI

Confidence interval

Authors’ contributions

SH, RL and NH contributed to the study design, data acquisition and analysis, interpretation of results, and writing and revising the manuscript. YW and QL contributed to the study design, interpretation of results, and reviewing the manuscript. JG contributed to data acquisition, data analysis, and reviewing of the manuscript. HX contributed to the study design, interpretation of results, and reviewing the manuscript. JB revised the manuscript. All authors read and approved the final manuscript.

Funding

The National Natural Science Foundation of China (82560239), The Key Incubation Project Funds of the second hospital & clinical medical school, lanzhou university (2025–25-zdfy-012).

Data availability

This study is based on data provided by Nationwide Inpatient Sample (NIS) database, part of the Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality. The NIS database is a large publicly available full-payer inpatient care database in the United States and the direct web link to the database is https://www.ahrq.gov/data/hcup/index.html. Therefore, individual or grouped data cannot be shared by the authors.

Declarations

Ethics approval and consent to participate

Not applicable. Administrative permissions were required to access the raw data used in this study and the author's (Hao Xie) work unit (Division of Orthopaedic Surgery, Department of Orthopaedics, Nanfang Hospital, Southern Medical University, Guangzhou, Guangdong, 510515, China.) has already granted permission from Agency for Healthcare Research and Quality (AHRQ) to access Healthcare Cost and Utilization Project (HCUP) Nationwide Databases. However, this observational study used deidentifed publicly available data, hence there was no requirement for consent to participate and it was deemed exempt by the ethics committee. So there is no need to grant permission in the Ethics approval and consent to participate section. What is more, the data used in this study were no need anonymized before its use. All methods are carried out following relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.