Abstract
Background
Children with a respiratory disease requiring invasive mechanical ventilation (IMV) in the paediatric intensive care unit (PICU) have an elevated risk for subsequent neurodevelopmental and behavioural disorders (NDBD). This study evaluates NDBD in children receiving IMV during surgical admissions.
Methods
Children enrolled in Texas Medicaid between 1999 and 2012 with a surgical admission were evaluated. Children in the PICU receiving IMV, in the PICU not receiving IMV, and in the intermediate medical care unit (IMCU) were identified and matched to children admitted to the general ward. The primary outcome was post-discharge NDBD. Secondary analyses evaluated NDBD risk by IMV duration and post-discharge psychotropic medication use.
Results
Of 35 161 children with surgical admissions meeting eligibility criteria, 993 were in the PICU with IMV, 7670 in the PICU without IMV, and 1027 in the IMCU. Increased rates of NDBD were observed in children receiving IMV (hazard ratio [HR] 1.91, 95% confidence interval [CI] 1.27–2.89, P=0.002), but not in those in the PICU without IMV (HR 1.12, 95% CI 0.98–1.29, P=0.10) or IMCU (HR 0.88, 95% CI 0.61–1.26, P=0.48). Elevated rates of NDBD were detected primarily in children receiving IMV for 96 h or more. Increased psychotropic medication use was observed only in the IMV group.
Conclusions
Children receiving invasive mechanical ventilation during a surgical admission are at increased risk of neurodevelopmental and behavioural disorders after hospital discharge. Further research is needed to clarify the mechanisms behind this association and to identify potentially modifiable risk factors.
Keywords: behavioural disorders, child neurodevelopment, critical care medicine, mechanical ventilation, paediatric critical care, surgical critical care
Editor's key points.
-
•
Mechanical ventilation for respiratory failure has been associated with higher rates of subsequent neurodevelopmental disorders and psychotropic medication use.
-
•
This study assessed whether surgery coupled with perioperative invasive mechanical ventilation in the ICU is associated with a greater incidence of neurodevelopmental and behavioural disorders beyond surgery alone.
-
•
Of 35 161 children in a US Medicaid database meeting eligibility criteria, 993 required mechanical ventilation in the ICU, which was associated with an increased risk of neurodevelopmental and behavioural disorders compared with those not requiring mechanical ventilation.
-
•
Further studies are required to determine whether there is a causal link, but children requiring perioperative mechanical ventilation might benefit from closer neuropsychological surveillance after their surgery.
Invasive mechanical ventilation (IMV) is used frequently in the paediatric intensive care unit (PICU) for a variety of critically ill patients, including children with respiratory failure, at risk for airway compromise, or requiring airway protection. Although IMV is a lifesaving procedure, its use for respiratory failure has been associated with higher rates of subsequent neurodevelopmental disorders and psychotropic medication use,1 and with lower scores in IQ, nonverbal memory, visuospatial skills, and fine motor function.2,3 Evaluating IMV is complex, as it is associated with factors such as hypoxaemia, infection, inflammation,4, 5, 6 sedative and analgesic medications,7, 8, 9, 10 and delirium,11, 12, 13 which have all been associated with neurodevelopmental deficits. Because studies of IMV in children most commonly evaluate those with acute respiratory failure resulting from respiratory illness, indication bias may confound the association between IMV and worse neurodevelopmental outcomes. This bias can be particularly problematic among children with baseline comorbid conditions predisposing to both acute respiratory illness and differences in neurodevelopmental trajectory.
We evaluated whether children receiving IMV during a surgical admission, an indication different from admission for severe respiratory disease, have a similar increased risk of neurodevelopmental and behavioural disorders (NDBD). In addition, given that prior studies have reported that early exposure to general anaesthesia for surgery is associated with an increased risk of behavioural and learning deficits,10 the goal of the current study was to assess whether surgery coupled with the necessity for perioperative IMV in the PICU is associated with an elevated incidence of NDMD beyond surgery alone. This was accomplished by comparing children with surgical procedures requiring perioperative IMV with those receiving procedures that did not require IMV.
Methods
This study was approved by the Institutional Review Board at Columbia University Irving Medical Center (New York, NY, USA) with a waiver of written informed consent. The manuscript follows the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines.14
Cohort generation
The Centers for Medicare and Medicaid Services (CMS) Medicaid Analytic eXtract (MAX) files for the years 1999–2012 were evaluated, with claims for all children enrolled in Texas Medicaid during that time linked to create a longitudinal dataset. Texas Medicaid claims were used because of the presence of variables identifying admission to the intensive care unit (ICU), use of mechanical ventilation, and presence of long-term neuropsychological diagnoses. Surgical procedures were identified using Healthcare Cost and Utilization Project (HCUP) surgery flag software based on ICD-9 and CPT codes for therapeutic or diagnostic procedures requiring an operating room.15
Children aged 28 days to 18 yr with a surgical procedure during inpatient hospitalisation were included in the study. Hospitalisations meeting the following criteria were excluded: neonatal admissions; admissions involving the psychiatric ICU; admissions with discharge to hospice, psychiatric facility, or unclear location; admissions that ended in child death or involved IMV outside the ICU;16,17 and hospitalisations with missing data concerning admission or discharge dates.
We ensured adequate assessment of baseline comorbid conditions by requiring a minimum period of Medicaid eligibility before inpatient surgical admission. Children aged 1 yr or older required Medicaid eligibility for at least 6 of the 12 months before the surgical admission, and children <1 yr old required eligibility for at least half of their life. To limit the number of children with significant baseline comorbid disease likely to alter developmental trajectory, children were excluded if they had a newborn hospitalisation with a length of stay >7 days, >1 admission to a general inpatient unit before the index surgical admission, or any hospitalisation before the surgical admission in question involving surgery or an ICU stay.1 Children with technology-dependent complex clinical conditions (CCCs) were excluded.18 Children who met the criteria for any primary or secondary outcome (below) before the index surgical admission were excluded. As two outpatient claims or one claim from an inpatient hospitalisation were required for exclusion based on the presence of a prior psychiatric diagnosis, children with a single prior outpatient claim remained in the study cohort. Children with an unknown biological sex were excluded.
Exposure and covariate selection
Children were classified into three exposure groups based on IMV use and hospital location during the surgical admission: (1) admission to the PICU requiring IMV; (2) admission to the PICU but not requiring IMV; and (3) admission to the intermediate medical care unit (IMCU). Children admitted only to the general inpatient unit served as the reference group. IMV outside the operating room was identified using ICD-9 procedure code 96.7X (continuous IMV), and in secondary analyses, it was further divided based on the duration of IMV into 96.70 (unspecified), 96.71 (fewer than 96 h), and 96.72 (96 h or more). Locations of children during their inpatient admission were identified using UB92 accommodation codes. PICU admissions were identified by codes beginning with 020 (intensive care) and 021 (coronary care unit). Codes 206 or 214 identified admission to the IMCU. Code 0204 identified admissions to the psychiatric ICU, an exclusion criterion. All remaining children were considered general inpatient admissions.
Potential confounders included sociodemographic covariates (age, year of birth, sex, race, ethnicity, and median income of ZIP code of residence), resource utilisation (prior hospitalisation or outpatient surgery, emergency department visits, and outpatient visits), clinical diagnoses (presence of CCCs, HCUP non-chronic and chronic conditions, and perinatal/gestational conditions),18,19 and surgery type. Surgery type was defined using the Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project Clinical Classifications Software (CCS).20 This software tool classified surgical procedures by organ system based on ICD-9 codes present on the hospital claims. In select cases, leading zeros were missing from procedure codes in the Texas MAX data, so the CCS classified the procedure into two different organ systems. In these situations, two authors (OI and CI) independently selected the more appropriate procedure based on the child's admission diagnosis. Conflicting selections were resolved by consultation with a third reviewing author (AG).
Outcomes
The primary outcome was a new ICD-9-coded NDBD after hospital discharge. This composite outcome included diagnoses of ASD, ADHD, learning difficulty, developmental coordination disorders, speech and language disorders, intellectual disability, or behavioural disorders, identified using claims-based algorithms validated as having high positive predictive values,21 or diagnoses of anxiety, depression, bipolar disorder, or other mood disorders (Supplementary Table S1).1,21, 22, 23 In addition, we evaluated the presence of persistent psychotropic medication use, defined as at least two filled 30-day prescriptions. Prescriptions included five classes: ADHD medications, sedative-anxiolytics, antidepressants, antipsychotics, and mood stabilisers (Supplementary Table S2).
Statistical analysis
Children in the general inpatient group were matched one-to-one without replacement to children in the three exposure groups using propensity scores estimated by multinomial generalised boosted regression modelling.24 Propensity scores were conditional on all covariates except surgery type. The maximum calliper distance for matching was 0.2.25 Covariate balance was evaluated before and after matching by calculating the standardised mean difference (SMD) for each covariate. A post-matching SMD >0.1 for any covariate was considered unbalanced, stricter than the traditional 0.2 standard.26,27 Surgery type and any unbalanced covariate with a post-match SMD >0.1 were also adjusted for by inclusion in outcome regression models.28 To evaluate the primary outcome of NDBD, Cox proportional hazard models, adjusting for surgery type and unbalanced covariates with SMD > 0.1, calculated hazard ratios (HRs) and 95% confidence intervals (CIs) between each exposure group and its matched controls.
Surgery type was included as a covariate instead of being matched in the primary analysis because given the close association between surgery type and postoperative disposition, matching on that covariate considerably reduced sample size, and, in addition to reducing study power, might introduce selection bias and limit external validity. To address this and other issues, three sensitivity analyses were performed. Firstly, to assess for residual confounding by not including surgery type in the propensity score match, we performed an analysis in which surgery type was also included in the matching algorithm. Secondly, to mitigate surveillance bias, another sensitivity analysis repeated the primary analysis but excluded NDBD within 90 days of hospital discharge. Thirdly, to assess the potential influence of surgical invasiveness, the primary analysis was repeated, including whether the child received a blood transfusion during the admission in outcome regression models.
Several secondary analyses were performed. To evaluate whether longer exposures to IMV exhibit a stronger association with subsequent NDBD, children in the PICU receiving IMV were stratified by IMV duration as >96 h vs <96 h. It should be noted that specific procedures can require longer exposures to IMV. To directly compare children in the PICU receiving IMV with those in the PICU without IMV, children in these two exposure groups were evaluated after matching on all covariates except surgery type, with surgery type adjusted for in outcome models. Children were then evaluated after matching on all covariates including surgery type. Psychotropic medication use after discharge was also evaluated using the primary matched cohort, estimating HRs with 95% CIs for the three exposure groups. Many mood stabilisers are also commonly used as anti-epileptic medications, and it is not possible to know for which indication these medications were prescribed. We performed a sensitivity analysis in which mood stabilisers were excluded from the list of potential psychotropic medications.
Cohort generation was completed using SAS v9.4 (SAS Institute, Cary, NC, USA).29 Propensity score matching and statistical analysis were performed using R v4.2.1 (R Foundation for Statistical Computing, Vienna, Austria).30 Propensity scores were estimated using the twang package, and the matching process and balance evaluation were performed using the Matchit and Cobalt packages, respectively.31, 32, 33
Results
We identified 74 652 children in the Texas Medicaid dataset between 1999 and 2012 with surgical admissions meeting inclusion criteria (Supplementary Table S3). After excluding children with inadequate Medicaid eligibility, prior psychiatric diagnoses, or major chronic conditions, 35 161 children with surgical hospitalisations remained (Fig. 1 and Supplementary Table S4). Of these children, 993 were admitted to the PICU with IMV, 7670 to the PICU without IMV, and 1027 to the IMCU. There were 25 471 potential control children in the general inpatient ward. Compared with those in the general inpatient ward, children in each exposure group had more resource utilisation and higher rates of comorbid conditions (Supplementary Table S5).
Fig 1.
Flowchart of children included for analysis. ICU, intensive care unit; IMCU, intermediate medical care unit; PICU, paediatric intensive care unit. ∗In compliance with the Medicaid Data Use Agreement, counts of 10 or fewer are suppressed.
Matching with general inpatients
We matched 890 (89.6%) children in the PICU with IMV group, 5828 (76.0%) children in the PICU without IMV group, and 965 (94.0%) children in the IMCU group to children in the general inpatient ward. Selected covariates between children in each exposure group and matched control children are displayed in Table 1, with a complete list detailed in Supplementary Table S6. Unmatched children were younger, had higher resource utilisation before admission, and had more comorbid medical conditions (Supplementary Table S6). After matching, all covariate SMDs were <0.2 in each exposure group, with nearly all SMDs <0.1 (Fig. 2).
Table 1.
Selected covariates for children by study group after matching against children admitted to the floor after surgery. ED, emergency department; IMCU, intermediate medical care unit; IMV, invasive mechanical ventilation; PICU, paediatric intensive care unit. ∗In this study, two outpatient claims or one inpatient claim were required to meet the exclusion criteria of prior psychiatric disease. These counts represent patients with only one outpatient claim related to psychiatric diagnoses before the surgical admission and were therefore not excluded. †In compliance with the Medicaid Data Use Agreement, counts of 10 or fewer are suppressed.
| IMV group (n=890) |
PICU group (n=5828) |
IMCU group (n=965) |
||||
|---|---|---|---|---|---|---|
| IMV | Floor | PICU | Floor | IMCU | Floor | |
| Age (months), median (IQR) | 12 (4–40) | 10 (3–37) | 25 (7–86) | 23 (7–85) | 31 (8–103) | 28 (8–106) |
| Male sex | 521 (58.5) | 532 (59.8) | 3250 (55.8) | 3212 (55.1) | 572 (59.3) | 599 (62.1) |
| Prior ED visit | 508 (57.1) | 492 (55.3) | 3683 (63.2) | 3709 (63.6) | 624 (64.7) | 631 (65.4) |
| Prior inpatient hospitalisation | 178 (20.0) | 204 (22.9) | 1121 (19.2) | 1116 (19.1) | 196 (20.3) | 192 (19.9) |
| Prior outpatient surgery | 67 (7.5) | 70 (7.9) | 548 (9.4) | 526 (9.0) | 110 (11.4) | 134 (13.9) |
| Non-chronic clinical conditions | ||||||
| Circulatory | 13 (1.5) | 19 (2.1) | 82 (1.4) | 88 (1.5) | 14 (1.5) | 21 (2.2) |
| Psychiatric∗ | —† (0.0) | — (0.0) | — (0.1) | — (0.1) | — (0.0) | — (0.0) |
| Neurologic | 402 (45.2) | 390 (43.8) | 3027 (51.9) | 2976 (51.1) | 563 (58.3) | 544 (56.4) |
| Perinatal | 347 (39.0) | 371 (41.7) | 1921 (33.0) | 1919 (32.9) | 322 (33.4) | 302 (31.3) |
| Respiratory | 595 (66.9) | 595 (66.9) | 3917 (67.2) | 3902 (67.0) | 671 (69.5) | 668 (69.2) |
| Chronic clinical conditions | ||||||
| Cardiovascular | 96 (10.8) | 123 (13.8) | 642 (11.0) | 704 (12.1) | 86 (8.9) | 124 (12.8) |
| Congenital | 293 (32.9) | 293 (32.9) | 2802 (48.1) | 2802 (48.1) | 410 (42.5) | 410 (42.5) |
| Psychiatric∗ | 73 (8.2) | 75 (8.4) | 528 (9.1) | 552 (9.5) | 105 (10.9) | 98 (10.2) |
| Neurologic | 181 (20.3) | 183 (20.6) | 1515 (26.0) | 1533 (26.3) | 257 (26.6) | 277 (28.7) |
| Perinatal | — (0.3) | — (0.1) | 25 (0.4) | 23 (0.4) | — (0.4) | — (0.5) |
| Respiratory | 231 (26.0) | 229 (25.7) | 1689 (29.0) | 1675 (28.7) | 327 (33.9) | 351 (36.4) |
| Complex chronic conditions | ||||||
| Cardiovascular | 79 (8.9) | 118 (13.3) | 407 (7.0) | 498 (8.5) | 51 (5.3) | 82 (8.5) |
| Congenital and genetic | 61 (6.9) | 78 (8.8) | 518 (8.9) | 606 (10.4) | 75 (7.8) | 91 (9.4) |
| Neonatal | 32 (3.6) | 43 (4.8) | 121 (2.1) | 128 (2.2) | 12 (1.2) | — (0.8) |
| Respiratory | 17 (1.9) | 26 (2.9) | 97 (1.7) | 100 (1.7) | 20 (2.1) | 24 (2.5) |
| Surgery type | ||||||
| Neurological | 213 (23.9) | 68 (7.6) | 1413 (24.2) | 582 (10.0) | 145 (15.0) | 95 (9.8) |
| General surgery | 208 (23.4) | 355 (39.9) | 1207 (20.7) | 2024 (34.7) | 277 (28.7) | 368 (38.1) |
| Ophthalmology and Otolaryngology | 38 (4.3) | 118 (13.3) | 351 (6.0) | 764 (13.1) | 106 (11.0) | 117 (12.1) |
| Respiratory | 119 (13.4) | 11 (1.2) | 404 (6.9) | 92 (1.6) | 45 (4.7) | 13 (1.3) |
| Cardiovascular | 172 (19.3) | 70 (7.9) | 1231 (21.1) | 352 (6.0) | 105 (10.9) | 58 (6.0) |
| Miscellaneous | — (0.7) | — (1.1) | 64 (1.1) | 50 (0.9) | — (1.0) | — (0.6) |
| Genitourinary | 12 (1.3) | 81 (9.1) | 210 (3.6) | 588 (10.1) | 42 (4.4) | 82 (8.5) |
| Orthopaedic and musculoskeletal | 80 (9.0) | 137 (15.4) | 722 (12.4) | 1109 (19.0) | 143 (14.8) | 176 (18.2) |
| Dermatologic | 42 (4.7) | 40 (4.5) | 226 (3.9) | 267 (4.6) | 92 (9.5) | 50 (5.2) |
Fig 2.
Standardised mean differences before and after primary match in exposure groups relative to matched controls. IMCU, intermediate medical care unit; IMV, invasive mechanical ventilation; PICU, paediatric intensive care unit.
Neurodevelopmental and behavioural disorders after paediatric surgical admissions
Figure 3 displays cumulative probabilities of NDBD for children in each exposure group compared with their matched controls, with censoring at loss of Medicaid eligibility. Probabilities of NDBD in children in the PICU with IMV group diverged from that of their corresponding controls following hospital discharge. The incidence of individual NDBD for each exposure group is included in Supplementary Table S7; the most common diagnoses were speech and language disorders, intellectual delay, developmental coordination disorder, and ADHD. After matching and adjustment for unbalanced covariates and surgical type, children admitted to the PICU with IMV had a higher rate of NDBD than matched controls (HR 1.91, 95% CI 1.27–2.89, P=0.002). No increased risk was seen in the PICU without IMV (HR 1.12, 95% CI 0.98–1.29, P=0.10) or IMCU (HR 0.88, 95% CI 0.61–1.26, P=0.48) groups (Fig. 4).
Fig 3.
Cumulative probabilities of neurodevelopmental and behavioural disorders in primary analysis for each exposure group. Includes the number of subjects at risk at various time points after discharge in the exposed and matched control children and 95% confidence bands around the cumulative probabilities. IMCU, intermediate medical care unit; IMV, invasive mechanical ventilation; PICU, paediatric intensive care unit.
Fig 4.
HRs of composite neurodevelopmental and behavioural disorders for each exposure group. CI, confidence interval; HR, hazard ratio; IMCU, intermediate medical care unit; IMV, invasive mechanical ventilation; PICU, paediatric intensive care unit. ∗Adjusting for surgery type and all unbalanced covariates in outcome model, but not matching on surgery type.
In the sensitivity analysis matching on surgery type, an increased risk of NDBD was observed in the PICU with and without IMV groups, but not in the IMCU group (Fig. 4). Post-match covariates by exposure group and SMDs after matching are displayed in Supplementary Table S8 and Supplementary Figure S1, respectively. In the sensitivity analysis repeating the primary analysis but excluding children with diagnoses within 90 days, only the PICU with IMV group carried an increased risk of NDBD (Fig. 4). Children admitted to the PICU were more likely to receive blood transfusions (Supplementary Table S5), but similar results were observed after also adjusting for perioperative blood transfusion as a proxy for surgical invasiveness (Fig. 4).
When assessing the influence of duration of IMV, of 890 matched children in the IMV group, 434 (48.8%) received IMV for 96 h or more, 453 (50.9%) received IMV for < 96 h, and 3 (0.3%) had an unspecified duration and were excluded from the analysis. After adjustment for surgery type and unbalanced covariates, children receiving IMV for at least 96 h had an increased risk of NDBD (HR 3.02, 95% CI 1.57–5.79, P<0.001) relative to matched children admitted to the ward, whereas children receiving IMV for <96 h did not (HR 1.32, 95% CI 0.73–2.38, P=0.35).
In the direct comparison between children in the PICU who did and did not receive IMV, 990 (99.7%) children receiving IMV were matched, and all covariate SMDs were <0.1 (Supplementary Table S9). Relative to children in the PICU without IMV, those receiving IMV had an increased risk of NDBD (HR 1.65, 95% CI 1.23–2.23, P<0.001). When also matching on surgery type, 989 (99.6%) children receiving IMV were matched, all covariate SMDs were <0.1, and similar results were observed (HR 1.56, 95% CI 1.17–2.08, P=0.002).
Psychotropic medication usage after discharge
Increased psychotropic medication use after discharge was seen in both children in the PICU receiving IMV (HR 3.22, 95% CI 1.15–9.03, P=0.03) and children in the PICU not requiring IMV (HR 1.32, 95% CI 1.02–1.70, P=0.03) relative to matched general hospital controls. This relationship was not observed in children in the IMCU (Fig. 5). After excluding possible anti-epileptic drugs, statistically significantly increased hazard was only seen in the PICU with IMV group (Fig. 5). Supplementary Table S10 details the incidence of prolonged post-discharge use of individual classes of psychotropic medications by exposure group.
Fig 5.
HRs of psychotropic medication use after discharge for each exposure group. CI, confidence interval; HR, hazard ratio; IMCU, intermediate medical care unit; IMV, invasive mechanical ventilation; PICU, paediatric intensive care unit.
Discussion
In this study of Medicaid-enrolled children without major preceding hospitalisation or technology dependence, children receiving IMV in the PICU during a surgical admission had a 1.9-fold increased risk of developing an NDBD after discharge compared with children admitted to the general ward. This association was consistent across techniques of adjustment for procedure type and surgical invasiveness, continued to occur at increased rates more than 3 months after discharge, and was stronger in children receiving IMV for 96 h or more. Children in the PICU not requiring IMV did not have significantly elevated rates of NDBD relative to children admitted to the general inpatient ward except in the secondary analysis matching on surgery type. However, the effect estimates and CIs were similar, suggesting that the results are similar between all adjustment methods. Elevated rates of NDBD were also not seen in children admitted to the IMCU. In analyses directly comparing children in the PICU receiving IMV with those in the PICU who did not, children receiving IMV had a 65% increased rate of NDBD diagnoses. Finally, children receiving perioperative IMV were more likely to receive psychotropic medications after discharge, a result that remained consistent even after excluding drugs with anti-epileptic indications.
Multiple neurodevelopmental deficits have been identified in children after PICU admission, including decreased IQ,6 working memory,34 and attention and executive function35 and increased problems with behaviour and school performance.36 Deficits were found to persist even when children were re-evaluated 12 months after recovery.37 Although the majority of these studies evaluated children with sepsis or meningococcal septic shock, deficits were seen in subjects with non-neurological, noncardiac, and non-septic critical illness as well.6 Children specifically requiring IMV for acute respiratory failure have also been evaluated and have higher risk of subsequent mental disorder diagnoses compared with children admitted to the hospital with similar baseline characteristics1 and lower IQ scores than their unexposed siblings.2 Therefore the observed excess risk of NDBD in children requiring IMV during a surgical admission is consistent with prior studies evaluating children requiring IMV in the PICU for medical aetiologies including severe respiratory illness.1,2
There are a variety of reasons why children who receive IMV may have a higher rate of subsequent neurodevelopmental deficit. These factors include exposure to hypoxaemia, increased inflammation, sedative and analgesic medications, inadequate pain control, and prolonged hospitalisation and immobilisation, which can be consequences of the indicated surgery that the child received. In adult populations, these factors are associated with an increased risk of delirium and long-term cognitive dysfunction.5,7,8,11, 12, 13,38 In children, delirium has also been associated with cognitive dysfunction after discharge and overall mortality,39,40 and the use of sedative medications might be independently associated with adverse neurodevelopmental outcomes. Further research is required to determine whether there is a causal relationship between any of these factors associated with IMV and subsequent paediatric neurodevelopment and whether they are modifiable.
Limitations
This study has several limitations. Firstly, despite matching on a wide range of sociodemographic and clinical factors, as with any observational study, unmeasured confounding and confounding by indication might limit the interpretation of the reported association. Even after matching on baseline comorbidities and other covariates and adjusting for procedure type, children who receive IMV during a surgical admission might still have a higher risk of neurodevelopmental deficit for a variety of reasons independent of IMV, including higher baseline medical disease severity. Children requiring prolonged hospital admission at the time of birth or with ICU or multiple prior hospital admissions before surgery were excluded from analysis, removing children with the most severe congenital disorders. However, a significant proportion of the remaining children received cardiac and neurosurgical procedures, and these children might have a higher risk of neurodevelopmental problems after surgery. Although adjustment and matching for the organ system of the surgical procedure was performed, there can be imprecision in this adjustment method. A variety of procedures can be performed within each organ system, so accounting for surgery type does not perfectly balance the invasiveness of the procedure. The sensitivity analysis adjusting for perioperative blood transfusions attempts to further adjust for surgical invasiveness, but residual confounding is likely to persist. The finding that longer perioperative IMV is associated with increased NDBD identifies children at greatest risk. However, this exploratory analysis should not be interpreted as causal because it is likely confounded by several factors: greater illness severity, operative complications, surgical invasiveness, and underlying comorbidity also frequently require longer periods of IMV. In addition, inaccuracy in these billing codes may be present, although relative to children receiving <96 h, hospital length of stay was appropriately longer in children receiving at least 96 h suggesting it might accurately indicate longer IMV duration.
Secondly, the need for IMV for medical or surgical reasons is associated with a variety of factors including hypoxaemia, hypercapnia and hypocapnia, infection, inflammation, fluid shifts, sedative and analgesic medication use, immobility, and delirium. For surgery specifically, factors such as the invasiveness of the procedure, tissue trauma, and blood loss might also play a role. This administrative data source did not allow us to distinguish among contributions from these factors.
Thirdly, our matching algorithm also excluded the youngest children with the highest levels of pre-existing medical diagnoses receiving IMV in the PICU (Supplementary Table S6). These results might not generalise to that population. In addition, the cohort included children enrolled in Texas Medicaid from 1999 to 2012: a population of lower socioeconomic status from a specific geographic region receiving care in a period where clinical practices may differ from the present day. Although nearly 40% of children in the USA are covered by Medicaid,41 this association might differ in children of other sociodemographic strata, other parts of the world, and more contemporary practices.
Fourthly, given that neonatal admissions were excluded, the results from this study cannot be generalised to that population. Finally, a limitation of Texas Medicaid data is the inability to identify whether IMV was strictly postoperative, and children receiving IMV before surgery may have medical indications for IMV that are separate from their indication for surgery.
Conclusions
Children receiving invasive mechanical ventilation in the intensive care unit during surgical admissions had higher subsequent rates of neurodevelopmental and behavioural disorders than those not requiring IMV perioperatively. The risk appears greater in children receiving IMV for multiple days. This observational study does not directly imply causality, as unmeasured confounding factors are almost certain to be present, including the underlying condition necessitating surgery. Further research is warranted to verify these results and determine whether there is a causal link between IMV or factors associated with IMV and higher rates of neurodevelopmental disorders. However, these results suggest that children requiring perioperative IMV might benefit from closer neuropsychological surveillance in the years after their surgery.
Authors’ contributions
Study design and conceptualisation: all authors
Cohort generation: LG, AB-E, SG
Statistical analysis: OI, LG, CM, AG, CI
Interpretation of data: all authors
Drafting of the manuscript: OI, AG, CI
Critical revision of the manuscript: all authors
Approved the final manuscript as submitted and agree to be accountable for all aspects of the work: all authors
Funding
CI is supported by the US Agency for Healthcare Research and Quality (AHRQ) award R01HS026493. The content is solely the responsibility of the authors and does not necessarily represent the official views of the AHRQ. AG was supported during this study by the US National Center for Advancing Translational Sciences, National Institutes of Health, through grant KL2TR001874.
Declaration of interest
The authors declare no conflicts of interest.
Handling Editor: Hugh C Hemmings Jr
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.bja.2024.12.042.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
References
- 1.Geneslaw A.S., Lu Y., Miles C.H., et al. Long-term increases in mental disorder diagnoses after invasive mechanical ventilation for severe childhood respiratory disease: a propensity matched observational cohort study. Pediatr Crit Care Med. 2021;22:1013–1025. doi: 10.1097/PCC.0000000000002790. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Watson R.S., Beers S.R., Asaro L.A., et al. Association of acute respiratory failure in early childhood with long-term neurocognitive outcomes. JAMA. 2022;327:836–845. doi: 10.1001/jama.2022.1480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Yang C.F., Xue Y., Feng J.Y., Jia F.Y., Zhang Y., Li Y.M. Gross motor developmental dysfunctional outcomes in infantile and toddler pediatric intensive care unit survivors. BMC Pediatr. 2019;19:508. doi: 10.1186/s12887-019-1893-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Mikkelsen M.E., Christie J.D., Lanken P.N., et al. The adult respiratory distress syndrome cognitive outcomes study: long-term neuropsychological function in survivors of acute lung injury. Am J Respir Crit Care Med. 2012;185:1307–1315. doi: 10.1164/rccm.201111-2025OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Milbrandt E.B., Angus D.C. Potential mechanisms and markers of critical illness-associated cognitive dysfunction. Curr Opin Crit Care. 2005;11:355–359. doi: 10.1097/01.ccx.0000170508.63067.04. [DOI] [PubMed] [Google Scholar]
- 6.Kachmar A.G., Irving S.Y., Connolly C.A., Curley M.A.Q. A systematic review of risk factors associated with cognitive impairment after pediatric critical illness. Pediatr Crit Care Med. 2018;19:e164–e171. doi: 10.1097/PCC.0000000000001430. [DOI] [PubMed] [Google Scholar]
- 7.Kok L., Slooter A.J., Hillegers M.H., van Dijk D., Veldhuijzen D.S. Benzodiazepine use and neuropsychiatric outcomes in the ICU: a systematic review. Crit Care Med. 2018;46:1673–1680. doi: 10.1097/CCM.0000000000003300. [DOI] [PubMed] [Google Scholar]
- 8.Pandharipande P.P., Pun B.T., Herr D.L., et al. Effect of sedation with dexmedetomidine vs lorazepam on acute brain dysfunction in mechanically ventilated patients: the MENDS randomized controlled trial. JAMA. 2007;298:2644–2653. doi: 10.1001/jama.298.22.2644. [DOI] [PubMed] [Google Scholar]
- 9.van Zellem L., Utens E.M., de Wildt S.N., Vet N.J., Tibboel D., Buysse C. Analgesia-sedation in PICU and neurological outcome: a secondary analysis of long-term neuropsychological follow-up in meningococcal septic shock survivors. Pediatr Crit Care Med. 2014;15:189–196. doi: 10.1097/PCC.0000000000000044. [DOI] [PubMed] [Google Scholar]
- 10.Ing C., Jackson W.M., Zaccariello M.J., et al. Prospectively assessed neurodevelopmental outcomes in studies of anaesthetic neurotoxicity in children: a systematic review and meta-analysis. Br J Anaesth. 2021;126:433–444. doi: 10.1016/j.bja.2020.10.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Girard T.D., Jackson J.C., Pandharipande P.P., et al. Delirium as a predictor of long-term cognitive impairment in survivors of critical illness. Crit Care Med. 2010;38:1513–1520. doi: 10.1097/CCM.0b013e3181e47be1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Girard T.D., Thompson J.L., Pandharipande P.P., et al. Clinical phenotypes of delirium during critical illness and severity of subsequent long-term cognitive impairment: a prospective cohort study. Lancet Respir Med. 2018;6:213–222. doi: 10.1016/S2213-2600(18)30062-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Pandharipande P.P., Girard T.D., Jackson J.C., et al. Long-term cognitive impairment after critical illness. N Engl J Med. 2013;369:1306–1316. doi: 10.1056/NEJMoa1301372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.von Elm E., Altman D.G., Egger M., Pocock S.J., Gøtzsche P.C., Vandenbroucke J.P. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453–1457. doi: 10.1016/S0140-6736(07)61602-X. [DOI] [PubMed] [Google Scholar]
- 15.Healthcare Cost and utilization Project (HCUP) surgery flag software. Agency for Healthcare Research and Quality; 2014. https://www.hcup-us.ahrq.gov/toolssoftware/surgflags/surgeryflags.jsp Available from: [PubMed] [Google Scholar]
- 16.Sobotka S.A., Lynch E., Liao C., Graham R.J., Msall M.E. Autism and neurodevelopmental disability risks in children with tracheostomies and ventilators. Pediatr Pulmonol. 2024;59:1380–1387. doi: 10.1002/ppul.26921. [DOI] [PubMed] [Google Scholar]
- 17.Sobotka S.A., Lynch E., Msall M.E., Graham R.J. Early childhood developmental skills of children with tracheostomies and ventilators recently discharged home. Pediatr Pulmonol. 2023;58:853–865. doi: 10.1002/ppul.26265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Feudtner C., Feinstein J.A., Zhong W., Hall M., Dai D. Pediatric complex chronic conditions classification system version 2: updated for ICD-10 and complex medical technology dependence and transplantation. BMC Pediatr. 2014;14:199. doi: 10.1186/1471-2431-14-199. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Healthcare Cost and utilization Project (HCUP) chronic condition indicator (CCI) for ICD-9-CM. Agency for Healthcare Research and Quality, Rockville, MD. Available from: https://www.hcup-us.ahrq.gov/toolssoftware/chronic/chronic.jsp (accessed 1 December 2015).
- 20.HCUP Clinical Classifications Software (CCS) for ICD-9-CM. Healthcare Cost and utilization Project (HCUP). Agency for Healthcare Research and Quality, Rockville, MD. Available from: https://hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp (accessed 28 April 2022).
- 21.Straub L., Bateman B.T., Hernandez-Diaz S., et al. Validity of claims-based algorithms to identify neurodevelopmental disorders in children. Pharmacoepidemiol Drug Saf. 2021;30:1635–1642. doi: 10.1002/pds.5369. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Ing C., Sun M., Olfson M., et al. Age at exposure to surgery and anesthesia in children and association with mental disorder diagnosis. Anesth Analg. 2017;125:1988–1998. doi: 10.1213/ANE.0000000000002423. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Deo A.J., Castro V.M., Baker A., et al. Validation of an ICD-code-based case definition for psychotic illness across three health systems. Schizophr Bull. 2024;50:1482–1488. doi: 10.1093/schbul/sbae064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.McCaffrey D.F., Griffin B.A., Almirall D., Slaughter M.E., Ramchand R., Burgette L.F. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med. 2013;32:3388–3414. doi: 10.1002/sim.5753. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Austin P.C. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat. 2011;10:150–161. doi: 10.1002/pst.433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Rosenbaum P.R., Rubin D.B. Constructing a control-group using multivariate matched sampling methods that incorporate the propensity score. Am Stat. 1985;39:33–38. [Google Scholar]
- 27.Cochran W.G., Rubin D.B. Controlling bias in observational studies. Sankhya Ser A. 1973;35:417–446. [Google Scholar]
- 28.Rubin D.B., Thomas N. Combining propensity score matching with additional adjustments for prognostic covariates. J Am Stat Assoc. 2000;95:573–585. [Google Scholar]
- 29.The SAS system for Windows [computer program] SAS Inst.; Cary, NC: 2011. https://support.sas.com/software/94/ [Google Scholar]
- 30.R: a language and environment for statistical computing [computer program]. Version 4.2.1. R Foundation for Statistical Computing; Vienna, Austria: 2010. https://cran.r-project.org/ [Google Scholar]
- 31.twang: Toolkit for Weighting and Analysis of Nonequivalent Groups [computer program] R Package; 2021. https://cran.r-project.org/src/contrib/Archive/twang/?C=M;O=A [Google Scholar]
- 32.cobalt: Covariate Balance Tables and Plots [computer program] R Package; 2022. https://cran.r-project.org/src/contrib/Archive/cobalt/ [Google Scholar]
- 33.MatchIt: Nonparametric Preprocessing for Parametric Causal Inference [computer program] R Package; 2022. https://cran.r-project.org/src/contrib/Archive/MatchIt/?C=M;O=A [Google Scholar]
- 34.Elison S., Shears D., Nadel S., Sahakian B., Garralda M.E. Neuropsychological function in children following admission to paediatric intensive care: a pilot investigation. Intensive Care Med. 2008;34:1289–1293. doi: 10.1007/s00134-008-1093-9. [DOI] [PubMed] [Google Scholar]
- 35.Als L.C., Nadel S., Cooper M., Pierce C.M., Sahakian B.J., Garralda M.E. Neuropsychologic function three to six months following admission to the PICU with meningoencephalitis, sepsis, and other disorders: a prospective study of school-aged children. Crit Care Med. 2013;41:1094–1103. doi: 10.1097/CCM.0b013e318275d032. [DOI] [PubMed] [Google Scholar]
- 36.Kaur J., Singhi P., Singhi S., Malhi P., Saini A.G. Neurodevelopmental and behavioral outcomes in children with sepsis-associated encephalopathy admitted to pediatric intensive care unit: a prospective case control study. J Child Neurol. 2016;31:683–690. doi: 10.1177/0883073815610431. [DOI] [PubMed] [Google Scholar]
- 37.Als L.C., Tennant A., Nadel S., Cooper M., Pierce C.M., Garralda M.E. Persistence of neuropsychological deficits following pediatric critical illness. Crit Care Med. 2015;43:e312–e315. doi: 10.1097/CCM.0000000000001075. [DOI] [PubMed] [Google Scholar]
- 38.Patel B.K., Wolfe K.S., Patel S.B., et al. Effect of early mobilisation on long-term cognitive impairment in critical illness in the USA: a randomised controlled trial. Lancet Respir Med. 2023;11:563–572. doi: 10.1016/S2213-2600(22)00489-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Dervan L.A., Di Gennaro J.L., Farris R.W.D., Watson R.S. Delirium in a tertiary PICU: risk factors and outcomes. Pediatr Crit Care Med. 2020;21:21–32. doi: 10.1097/PCC.0000000000002126. [DOI] [PubMed] [Google Scholar]
- 40.Traube C., Silver G., Gerber L.M., et al. Delirium and mortality in critically ill children: epidemiology and outcomes of pediatric delirium. Crit Care Med. 2017;45:891–898. doi: 10.1097/CCM.0000000000002324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Health insurance coverage of children 0-18. KFF; 2022. https://www.kff.org/other/state-indicator/children-0-18/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D Available from: [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.