Outpatient insulin use in type 2 diabetes mellitus and acute respiratory distress syndrome outcomes: A retrospective cohort study

Abstract

BACKGROUND

The impact of type 2 diabetes mellitus (T2DM) on acute respiratory distress syndrome (ARDS) is debatable. T2DM was suspected to reduce the risk and complications of ARDS. However, during coronavirus disease 2019 (COVID-19), T2DM predisposed patients to ARDS, especially those who were on insulin at home.

AIM

To evaluate the impact of outpatient insulin use in T2DM patients on non-COVID-19 ARDS outcomes.

METHODS

We conducted a retrospective cohort analysis using the Nationwide Inpatient Sample database. Adult patients diagnosed with ARDS were stratified into insulin-dependent diabetes mellitus (DM) (IDDM) and non-insulin-dependent DM (NIDDM) groups. After applying exclusion criteria and matching over 20 variables, we compared cohorts for mortality, duration of mechanical ventilation, incidence of acute kidney injury (AKI), length of stay (LOS), hospitalization costs, and other clinical outcomes.

RESULTS

Following 1:1 propensity score matching, the analysis included 274 patients in each group. Notably, no statistically significant differences emerged between the IDDM and NIDDM groups in terms of mortality rates (32.8% vs 31.0%, P = 0.520), median hospital LOS (10 d, P = 0.537), requirement for mechanical ventilation, incidence rates of sepsis, pneumonia or AKI, median total hospitalization costs, or patient disposition upon discharge.

CONCLUSION

Compared to alternative anti-diabetic medications, outpatient insulin treatment does not appear to exert an independent influence on in-hospital morbidity or mortality in diabetic patients with non-COVID-19 ARDS.

Key Words: Acute respiratory distress syndrome; Type 2 diabetes mellitus; Insulin; Length of stay; Mortality; Endotracheal intubation; Acute kidney injury; Coronavirus disease 2019

Core Tip: During the coronavirus disease 2019 (COVID-19) era, multiple studies have shown that outpatient insulin use was a bad prognosticator for COVID-19-induced acute respiratory distress syndrome (ARDS). After the use of greedy propensity matching to balance multiple confounders, this study demonstrated that outpatient insulin use has no impact on non-COVID-19-induced ARDS compared to the use of other anti-hyperglycemic agents in type 2 diabetes patients. The outcomes evaluated were mortality, length of stay, duration of intubation, cost of stay, and acute kidney injury. These findings raise a lot of unanswered questions on the relationship between diabetes and acute lung injury outcomes.

INTRODUCTION

Acute respiratory distress syndrome (ARDS) is a severe pulmonary condition characterized by widespread lung inflammation, resulting in increased permeability of pulmonary capillaries, which causes interstitial and alveolar edema[1,2]. The clinical hallmark of ARDS was delineated by the Berlin Consensus Criteria, characterized by the acute onset of bilateral noncardiogenic pulmonary infiltrates visible on chest radiographs, accompanied by some degree of hypoxemia described as partial pressure of oxygen/fraction of inspired oxygen less than 300 mmHg with positive end-expiratory pressure or continuous positive airway pressure ≥ 5 cm H₂O, in the absence of left atrial hypertension[3]. Recognized for its elevated mortality rate of approximately 40%, ARDS often requires prolonged mechanical ventilation, leading to extended stays in the intensive care unit (ICU) and hospital. This pathology became a huge healthcare issue, especially during the coronavirus disease 2019 (COVID-19) era[2,4-6].

Studies have elucidated the pivotal role of comorbidities as risk modifiers and prognosticators in the context of ARDS[7,8]. Among these comorbidities, diabetes mellitus (DM) is very prevalent, representing a relationship subject to extensive research yielding diverse findings[9-11]. Some studies conclude that pre-existing DM increases ARDS risk especially in the setting of COVID-19[12]. In contrast, others suggest it does not significantly influence the incidence or mortality related to acute lung injury (ALI) or ARDS in critically ill patients[13]. Some even propose a potential reduction in risk, as indicated by a meta-analysis of seven cohort studies involving critically ill patients with common risk factors for ALI/ARDS[14]. This analysis concluded that diabetes was protective against the development of ARDS[14]. Consequently, DM is considered a protective factor in the "lung injury prediction score" used to prognosticate ALI risk in sepsis patients[7,8]. This protective effect might be explained by reduced microvascular protein extravasation[15] and lower concentrations of neutrophils, superoxides, and inflammatory cytokines in the bronchoalveolar fluid of diabetics[16-18].

Although there was a potential reduction in the risk of ARDS, it is important to consider mortality as there was no significant difference in mortality rates between ARDS individuals with DM and those without[11]. If decreased lung injury was responsible for the lower ARDS incidence, one would anticipate a parallel reduction in mortality. However, the lack of mortality benefit could be attributed to DM acting as an independent risk factor for mortality in critical illness and sepsis, which overshadows any potential mortality benefit attributed to reduced lung injury[11]. Moreover, DM might act as a confounding factor, where elements associated with treating or managing DM could influence the variable risk of developing lung injury. Therapies for DM may also alter any direct effects that DM could exert on ALI[19]. For instance, studies have demonstrated that insulin[20] exerts an anti-inflammatory effect by inducing vasodilation via release of endothelial nitric oxide in arteries, veins, and capillaries[12,13]. Additionally, it suppresses pro-inflammatory transcription factors in mononuclear cells[21,22], as well as intercellular cell adhesion molecular-1, MCP-1 expression, and NF-κB binding in human aortic endothelial cells in vitro[22-25]. Thus, it might be anticipated that diabetic patients on long-term insulin who develop ARDS, which is primarily an inflammatory condition, have better outcomes. However, the former was not the case in COVID-19-induced ARDS, as it was demonstrated that patients on outpatient insulin had worse outcomes then diabetic patients on oral anti-diabetics.

Building upon these insights, our primary objective was to examine the impact of home insulin use in patients with type 2 DM (T2DM) on outcomes of non-COVID-19 ARDS by distinguishing those with insulin-dependent DM (IDDM) and non-insulin-dependent DM (NIDDM). Our hypothesis posited that NIDDM patients who developed ARDS would experience worse outcomes compared to those with IDDM.

MATERIALS AND METHODS

Study design and database

This retrospective cohort analysis was conducted using the Nationwide Inpatient Sample (NIS) database for the year 2018, which constitutes the largest inpatient care database in the United States, encompassing data from over 7 million hospital visits. Institutional Review Board approval was deemed exempt under the Healthcare Cost and Utilization Project (HCUP) data use agreement (DUA) due to the de-identification of patient information. The research adhered to ethical guidelines set forth by the institution and HCUP-DUA.

Participants and eligibility criteria

Adult patients aged 18 years and older diagnosed with ARDS were included. Patients were categorized into two groups, based on their outpatient T2DM management: Those with IDDM and those with NIDDM. Patients were excluded if they had other types of diabetes (type 1 or secondary), were pregnant, had malignancies, exhibited a length of stay (LOS) less than 3 d, were trauma patients (due to insufficient data in the database), or were transplant patients. These criteria were chosen to focus the study on the specific impact of T2DM management on ARDS outcomes and to minimize potential confounding factors. Key data points included patient demographics (age, sex, race), comorbidities, etiologies of ARDS, and details of diabetes management.

Statistical analysis

Baseline characteristics were analyzed using descriptive statistics. Continuous variables were compared using independent t-test (for parametric data) or the Mann-Whitney U test (for nonparametric data). In contrast, categorical variables were compared using chi-square test or Fisher's exact test. We conducted propensity score matching at a 1:1 ratio to account for baseline differences between groups. This process was designed to balance over 20 baseline characteristics, including demographic factors and comorbidities such as smoking, chronic obstructive pulmonary disease, hypertension, dyslipidemia, chronic kidney disease, and cardiovascular disease, among others. A logistic regression model was employed to generate propensity scores for each patient, and a nearest-neighbor matching model with a caliper width of 0.1 was used for patient pairing.

After matching, the study aimed to assess clinical outcomes, including mortality, duration of mechanical ventilation, incidence of acute kidney injury (AKI), and other critical factors such as the hospital LOS and hospitalization costs.

All statistical analyses were performed using the Statistical Package for the Social Sciences (version 26; IBM Corp., Armonk, NY, United States) and R programming for matching procedures. Statistical significance was set at P less than 0.05.

RESULTS

Baseline characteristics

Out of 1174 patients diagnosed with ARDS, 274 had IDDM and 900 had NIDDM. The pre-match baseline characteristics of the patients are presented in Table 1. The mean age was comparable between groups, at 61.9 ± 13.3 years for IDDM and 63.5 ± 13.7 years for NIDDM (P = 0.731). Sex distribution was also similar, with females constituting 48.9% of the IDDM group and 48.4% of the NIDDM group (P = 0.894). However, a notable difference in race existed between the groups, with 19.7% of patients in the IDDM group being Black compared to only 13.9% in the NIDDM group (P = 0.019). Hypertension was more prevalent in the IDDM group (71.2%) compared to the NIDDM group (67.6%, P = 0.005), and dyslipidemia was significantly higher in the IDDM group (49.6% vs 38.8%, P = 0.001) (Table 1).

Table 1 Baseline characteristics of type 2 diabetes mellitus patients diagnosed with acute respiratory distress syndrome stratified into insulin-dependent and non-insulin-dependent groups (pre-match), n (%).

Characteristic	IDDM group, n = 274	NIDDM group, n = 900	P value
Age in yr	61.9 ± 13.3	63.5 ± 13.7	0.731
Female sex	134 (48.9)	436 (48.4)	0.894
Emergent	256 (93.4)	833 (92.7)	0.664
White	167 (60.9)	580 (64.4)	0.292
Black	54 (19.7)	125 (13.9)	0.019a
Hispanic	34 (12.4)	116 (12.9)	0.835
Income interval in USD
< 46000	106 (38.7)	329 (36.6)	0.523
46000-58999	81 (29.6)	241 (26.8)	0.366
59000-78999	47 (17.2)	207 (23.0)	0.040a
≥ 79000	40 (14.6)	123 (13.7)	0.696
Active smoking	45 (16.4)	143 (15.9)	0.833
COPD	68 (24.8)	253 (28.1)	0.284
CAD	78 (28.5)	217 (24.1)	0.146
CHF	128 (46.7)	406 (45.1)	0.641
CKD	78 (28.5)	247 (27.4)	0.740
HTN	195 (71.2)	608 (67.6)	0.005a
pHTN	29 (10.6)	76 (8.4)	0.277
CLD	30 (10.9)	116 (12.9)	0.394
ESRD	41 (15)	111 (12.3)	0.256
Asthma	20 (7.3)	56 (6.2)	0.526
Obesity	99 (36.1)	273 (30.3)	0.071
DLD	136 (49.6)	349 (38.8)	0.001a
AUD	10 (3.6)	54 (6.0)	0.134

^aP < 0.05.

AUD: Alcohol use disorder; CAD: Coronary artery disease; CKD: Chronic kidney disease; CHF: Congestive heart failure; CLD: Chronic liver disease; COPD: Chronic obstructive pulmonary disease; DLD: Dyslipidemia; ESRD: End-stage renal disease; HTN: Hypertension; IDDM: Insulin-dependent type 2 diabetes mellitus; NIDDM: Non-insulin-dependent type 2 diabetes mellitus; pHTN: Pulmonary hypertension; USD: United States dollars.

Clinical outcomes and complications

As illustrated in Table 2, the primary etiological factors for ARDS identified in the pre-matching phase encompassed pneumonia and sepsis. Pneumonia was present in 59.5% of patients with IDDM compared to 53.2% of patients with NIDDM (P = 0.068). Conversely, sepsis exhibited a higher prevalence in the NIDDM group (67.2%) compared to the IDDM group (58.4%, P = 0.007).

Table 2 Etiologies of acute respiratory distress syndrome before propensity score matching, n (%).

Etiology	IDDM group, n = 274	NIDDM group, n = 900	P value
Sepsis	160 (58.4)	605 (67.2)	0.007a
Transfusion	0 (0.0)	5 (0.6)	0.216
Pancreatitis	7 (2.6)	31 (3.4)	0.466
Surgery	2 (0.7)	20 (2.2)	0.111
Pneumonia	163 (59.5)	479 (53.2)	0.068a
Aspiration	55 (20.1)	220 (24.4)	0.135

^aP < 0.05.

IDDM: Insulin-dependent type 2 diabetes mellitus; NIDDM: Non-insulin-dependent type 2 diabetes mellitus.

Following propensity score matching and accounting for key variables such as age, sex, and comorbidities, each group consisted of 274 patients (Table 3). The prevalence of pneumonia as an etiological factor for ARDS remained comparable in both groups, with rates of 59.5% in the IDDM group and 58.8% in the NIDDM group (P = 0.862) (Table 4). Sepsis rates were also balanced post-matching, at 58.4% in the IDDM group and 61.7% in the NIDDM group (P = 0.433).

Table 3 Baseline characteristics of type 2 diabetes mellitus patients diagnosed with acute respiratory distress syndrome stratified into insulin-dependent and non-insulin-dependent groups (post-match), n (%).

Characteristic	IDDM groupa, n = 274	NIDDM groupa, n = 274	P value
Age in yr	61.9 ± 13.3	61.2 ± 15.1	0.117
Female sex	134 (48.9)	143 (52.2)	0.442
Emergent	256 (93.4)	253 (92.3)	0.618
White	167 (60.9)	163 (59.5)	0.727
Black	54 (19.7)	55 (20.1)	0.915
Hispanic	34 (12.4)	35 (12.8)	0.898
Income interval in USD
< 46000	106 (38.7)	109 (39.8)	0.793
46000-58999	81 (29.6)	79 (28.8)	0.851
59000-78999	47 (17.2)	39 (14.2)	0.347
≥ 79000	40 (14.6)	47 (17.2)	0.413
Active smoking	45 (16.4)	47 (17.2)	0.819
COPD	68 (24.8)	67 (24.5)	0.921
CAD	78 (28.5)	67 (24.5)	0.287
CHF	128 (46.7)	116 (42.3)	0.302
CKD	78 (28.5)	73 (26.6)	0.633
HTN	195 (71.2)	194 (70.8)	0.925
pHTN	29 (10.6)	25 (9.1)	0.566
CLD	30 (10.9)	36 (13.1)	0.431
ESRD	41 (15)	36 (13.1)	0.539
Asthma	20 (7.3)	28 (10.2)	0.227
Obesity	99 (36.1)	110 (40.1)	0.333
DLD	136 (49.6)	141 (51.5)	0.669
AUD	10 (3.6)	15 (5.5)	0.306

^aP < 0.05.

AUD: Alcohol use disorder; CAD: Coronary artery disease; CKD: Chronic kidney disease; CHF: Congestive heart failure; CLD: Chronic liver disease; COPD: Chronic obstructive pulmonary disease; DLD: Dyslipidemia; ESRD: End-stage renal disease; HTN: Hypertension; IDDM: Insulin-dependent type 2 diabetes mellitus; NIDDM: Non-insulin-dependent type 2 diabetes mellitus; pHTN: Pulmonary hypertension; USD: United States dollars.

Table 4 Etiologies of acute respiratory distress syndrome after propensity score matching, n (%).

Etiology	IDDM group, n = 274	NIDDM group, n = 274	P value
Sepsis	160 (58.4)	169 (61.7)	0.433
Transfusion	0 (0.0)	0 (0.0)	N/A
Pancreatitis	7 (2.6)	8 (2.9)	0.793
Surgery	2(0.7)	3 (1.1)	0.653
Pneumonia	163 (59.5)	161 (58.8)	0.862
Aspiration	55 (20.1)	50 (18.2)	0.587

IDDM: Insulin-dependent type 2 diabetes mellitus; N/A: Not applicable; NIDDM: Non-insulin-dependent type 2 diabetes mellitus.

Post-matching, the median LOS was identical for both groups, at 10 d (IDDM interquartile range: 5-19, NIDDM interquartile range: 5-17, P = 0.537), with comparable total hospital charges in United States dollars (USD) (130466 for the IDDM group and 134033 for the NIDDM group, P = 0.06). Additionally, there was no distinction in the disposition of patients upon discharge, whether to home, home with healthcare, or to a skilled nursing facility. However, among the 373 individuals discharged alive from the hospital, 271 (72.6%) required home services or admission to a skilled nursing facility (Table 5).

Table 5 Primary outcomes of acute respiratory distress syndrome assessed post-propensity score matching, n (%).

Outcome	IDDM group, n = 274	NIDDM group, n = 274	P value
AKI	140 (51.1)	146 (53.3)	0.600
Duration of mechanical ventilation in h
< 24	39 (14.2)	36 (13.1)	0.709
24-96	42 (15.3)	49 (17.9)	0.422
> 96	101 (36.9)	109 (39.8)	0.482
Mortality	90 (32.8)	85 (31.0)	0.520
Patient's disposition
Home	53 (19.3)	49 (17.9)	0.661
Home health care	28 (10.2)	28 (10.2)	0.999
Skilled nursing facility	103 (37.6)	112 (40.9)	0.431
Length of stay	10 (5-19)	10 (5-17)	0.537
Total charges in USD	130466 (59830-268939)	134033 (53771-272176)	0.060

AKI: Acute kidney injury; IDDM: Insulin-dependent type 2 diabetes mellitus; NIDDM: Non-insulin-dependent type 2 diabetes mellitus; USD: United States dollars.

Mortality rates were comparable, with 32.8% in the IDDM group and 31.0% in the NIDDM group (P = 0.520). Additionally, the IDDM group had a slightly higher rate of mechanical ventilation usage (42.3%) compared to the NIDDM group (39.7%), but this difference was not statistically significant (P = 0.451). Regarding complications such as the duration of mechanical ventilation, no significant difference was found between the groups. The incidence of AKI was higher in the NIDDM group (53.3%) than in the IDDM group (51.1%), though the difference was not statistically significant (P = 0.600) (Table 5).

DISCUSSION

To our knowledge, this is the first report of a retrospective cohort study analyzing the impact of outpatient management of T2DM on outcomes in non-COVID-19 ARDS patients. Specifically, we compared patients with IDDM to those with NIDDM. The study was designed as an analysis of a large, multicenter, prospective ICU patient dataset with national representation based on NIS. Rigorous selection criteria were paramount in ensuring the validity and reliability of our findings. By carefully defining selection criteria and considering factors such as age and specific comorbidities, we established a well-defined study population. Moreover, our methods encompassed propensity score matching, ensuring robust control for potential confounding variables and consequently improving the validity of our results, ensuring that the only variable impacting the outcomes was outpatient insulin use.

Using a sample from the population before COVID-19 (NIS 2018) emergence eliminates any confounding factors that might alter the results, such as COVID-19 severity or sequelae. Using ICD-10 codes for sampling ensured a standardized, comprehensive method for identifying relevant comorbidities, further enhancing the study’s methodological rigor. These selection criteria not only fortified the study's internal validity by reducing confounding variables but also aided in creating a more homogeneous and clinically relevant sample.

DM patients are at higher risk of developing ARDS due to predisposing conditions such as gastroparesis[26], immune system suppression, and autonomic neuropathy[27-29]. However, studies suggest that DM acts via an unclear mechanism as a risk modifier by reducing ARDS incidence in patients with predisposing conditions, such as sepsis[7,8,30,31]. Considering the well-documented anti-inflammatory effects of insulin, our objective aimed to discern if exogenous insulin use contributes to this protective effect, prompting an examination of the distinctions between IDDM and NIDDM.

In the initial examination of our hypotheses, two seemingly contradictory notions emerged. Firstly, it was posited that patients with IDDM might have more pronounced pancreatic aging and scarring, characteristic of long-standing advanced DM and insulin resistance. This condition could potentially lead to exacerbated hyperglycemia, oxidative stress, and systemic inflammation[20,32,33]. Consequently, such patients might experience worse outcomes in ARDS compared to diabetic individuals not utilizing home insulin. However, a counterargument was proposed based on the recognized anti-inflammatory properties of insulin[12,13,23-25], suggesting that insulin therapy could mitigate the adverse effects of hyperglycemia through metabolic regulation, thereby contributing to improved ARDS outcomes in diabetic patients. Our study yielded no significant differences in morbidity and mortality outcomes among T2DM patients on home insulin compared to those not utilizing insulin in the context of non-COVID-19 ARDS. Notably, this observation persisted even after matching for major ARDS causes, such as sepsis and pneumonia. The implications of this finding prompt a nuanced interpretation, suggesting that patients who do not require exogenous insulin may still benefit from residual natural insulin secretion, conferring some degree of protection. Moreover, patients on synthetic insulin may derive advantages from its anti-inflammatory properties. This finding aligns with prior investigations where DM was identified as independently protective against ARDS development even after adjusting for baseline parameters, including diabetic medications[30,31].

An important finding was that home insulin use for managing T2DM did not influence the duration of mechanical ventilation in ARDS patients requiring this intervention. This implies a comparable severity of lung inflammation necessitating prolonged ventilator support. This aligns with the hypothesis that both groups, regardless of home insulin use, exhibit the same degree of inflammation, suggesting that insulin does not confer an additional anti-inflammatory effect.

Regarding mortality rates, despite progress in understanding ARDS over the last 50 years, high mortality rates persist (21%-60%)[34-38] including among diabetic patients[36,39], aligning with our findings. However, the extent to which reported ARDS mortality can be attributed to the condition itself vs underlying comorbidities, such as cancer and immunosuppression, non-pulmonary organ dysfunction and older age, remains unclear[35].

Conversely, in COVID-19-related ARDS, a study involving 696 patients revealed that those with IDDM experience worse outcomes, including higher incidence of ARDS and mortality, compared to the NIDDM group[40]. Furthermore, both a case–control study[41] and a retrospective cohort study[42] indicated that diabetic patients on insulin, either as a sole treatment or in combination with other anti-diabetic agents, who contract COVID-19, exhibit a higher mortality rate compared to those not receiving insulin treatments. This observation was substantiated by a meta-analysis associating insulin use with increased COVID-19 mortality (odds ratio = 2.10, 95% confidence interval: 1.51-2.93), although the specific cause of death was not directly specified; however, lung inflammation and ARDS are commonly identified as primary causes of death in COVID-19[43]. These results were not implicated in our study, highlighting a potentially significant difference between COVID-19-induced ARDS and other ARDS subtypes.

The observed similarity in the incidence of AKI between both cohorts may be explained by a nuanced interplay of baseline renal risk and the mitigating effects of diabetes management strategies. IDDM patients typically have a more prolonged and possibly severe course of diabetes, which inherently increases their risk for renal complications, including AKI. However, the regular use of insulin, known for its anti-inflammatory properties and beneficial effects on endothelial function, might offer a protective shield, counterbalancing this heightened risk[20]. Conversely, NIDDM patients, generally managed with oral antidiabetics, likely have a lower baseline risk of renal impairment. However, they lack the potential renal protective benefits conferred by insulin, except for those SGLT-2 inhibitors and some GLP-1 agonists[44]. This dynamic could have resulted in a leveling of AKI risks across both groups despite the differing diabetes management approaches. This interpretation underscores the complex relationship between long-term diabetes management, inherent renal risk factors, and the acute pathophysiological challenges presented by ARDS, suggesting that the balance of risks and protective factors leads to similar AKI rates observed in both groups.

As for the administrative parameters influencing patients and the country's entire healthcare system such as the LOS, median total hospitalization costs, and patient disposition upon discharge, they were comparable between both groups.

The average LOS in hospitals denotes the average number of days patients spend in the hospital[45]. The average LOS for all hospitalizations in the United States in 2021 was 5.9 d[46]. In our study, the average LOS was 10 d, nearly double the total number of hospitalizations for all causes combined but within the reported range for the length of hospital stay for ARDS (9-30 d)[5,41,47-49]. This average LOS of 10 d resulted in an average cost of around 130000 USD per admission. Extrapolating this to the 1174 patients with T2DM diagnosed with ARDS during the same year would cost the United States’ healthcare system over 150 million USD. Furthermore, all other factors being equal, a shorter stay will reduce the cost per discharge and shift care from inpatient to less expensive post-acute settings[45,46]. Unfortunately, our findings showed that 271 of the 373 patients (72.6%) discharged alive from the hospital required home services or admission to a skilled nursing facility, which highlights the impactful socio-economic burden of ARDS.

There are several limitations in our study. Firstly, the reliance on the NIS database introduces the potential for coding inaccuracies. The use of ICD-10 codes for identifying patients with ARDS and IDDM/NIDDM, while standardized, does not allow for evaluating severity indicators such as hemoglobin A1c and home and in-hospital glucose levels. As with any database-dependent research, our results are dependent on the accuracy and completeness of the coding practices across various hospitals. Additionally, the retrospective nature of our study limits our ability to establish causation between ARDS and diabetes management and the observed morbidity/mortality outcomes. We identified associations, but the temporal sequence and causality remain speculative without longitudinal follow-up.

Furthermore, analyzing the specific causes of ARDS in patients proved challenging due to the overlapping nature of diagnoses. Many patients had multiple concurrent conditions coded under ICD-10, such as sepsis and pneumonia, which commonly coexist in the setting of critical illness. This overlap in diagnoses within the same admission made it difficult to attribute ARDS to a singular cause. As a result, the complexity of each patient's clinical picture posed a significant limitation in our ability to precisely determine an association between insulin use and the primary etiological factors.

It is worth noting that our study has some other limitations. We did not include detailed information about anti-hyperglycemic regimens like SGLT-2 inhibitors and GLP-1 agonists, which have shown promising outcomes. Additionally, we only focused on a single hospitalization period, precluding our ability to evaluate the long-term impact of treatment. Finally, our study was unable to determine the duration of diabetes.

CONCLUSION

In summary, our analysis suggests that the use of outpatient insulin, or lack thereof, in the treatment of T2DM does not appear to lead to significantly worse ARDS outcomes, in contrast to the findings in patients with COVID-19 ARDS. Moreover, insulin use does not contribute to the modifying effect observed in diabetes, and other factors may be at play. Consequently, further prospective studies are warranted to comprehensively explore these factors.

Future directions in research could involve the prospective evaluation of glycemic control targets for ARDS patients, focusing on linking glucose profiles to post-discharge outcomes. A deeper understanding of the mechanisms underlying these disparities and implementing equitable public health responses are essential steps toward addressing and mitigating these complexities.