Color bars show mean responses. AE indicates adverse effect; DNAR, do not attempt resuscitation; EA, esophageal atresia; LST, life-sustaining treatment; and PICC, peripherally inserted central catheter.
aSixteen clinical vignettes related to decisions faced by parents and physicians of infants in the neonatal intensive care unit were presented to each respondent. The vignettes were ordered top to bottom according to the value of the primary outcome variable of the preferred decision maker, from all parent to mostly parent, equally parent and medical team, mostly medical team, and all medical team.
bIndicates the preferred decision maker for each clinical vignette (primary outcome variable). Respondents were asked who the preferred decision maker was using the modified 5-point Control Preference Scale (parent- vs medical team–centered).
cSeven characteristics of decisions were assessed for each clinical vignette. For each question, the respondent chose a value between 1 (not at all) and 10 (very much) to indicate the extent to which that characteristic was present in each clinical vignette.
eMethods 1. Final Survey Vignettes
eMethods 2. Final Survey Questions
Customize your JAMA Network experience by selecting one or more topics from the list below.
Weiss EM, Xie D, Cook N, Coughlin K, Joffe S. Characteristics Associated With Preferences for Parent-Centered Decision Making in Neonatal Intensive Care. JAMA Pediatr. 2018;172(5):461–468. doi:10.1001/jamapediatrics.2017.5776
What characteristics of decisions are associated with preferences for parent- vs medical team–centered decision making among parents of children in the neonatal intensive care unit?
This cross-sectional survey of 136 parents of children in the neonatal intensive care unit showed that when decisions involved big-picture goals and had a high potential to harm the infant, parents tended to prefer parent-centered decision making. In contrast, when decisions were technical in nature, had a high potential for benefit, required medical expertise, and were urgent, parents tended to prefer medical team–centered decision making.
Decision-making preferences among parents of children in the neonatal intensive care unit may vary systematically by decision-level characteristics; incorporating this variation into shared decision making may improve the outcomes of medical decisions.
Little is known about how characteristics of particular clinical decisions influence decision-making preferences by patients or their surrogates. A better understanding of the factors underlying preferences is essential to improve the quality of shared decision making.
To identify the characteristics of particular decisions that are associated with parents’ preferences for family- vs medical team–centered decision making across the spectrum of clinical decisions that arise in the neonatal intensive care unit (NICU).
Design, Setting, and Participants
This cross-sectional survey assessed parents’ preferences for parent- vs medical team–centered decision making across 16 clinical decisions, along with parents’ assessments of 7 characteristics of those decisions. Respondents included 136 parents of infants in 1 of 3 academically affiliated hospital NICUs in Philadelphia, Pennsylvania, from January 7 to July 8, 2016. Respondents represented a wide range of educational levels, employment status, and household income but were predominantly female (109 [80.1%]), white (68 [50.0%]) or African American (53 [39.0%]), and married (81 of 132 responding [61.4%]).
Main Outcomes and Measures
Preferences for parent-centered decision making. For each decision characteristic (eg, urgency), multivariable analyses tested whether middle and high levels of that characteristic (compared with low levels) were associated with a preference for parent-centered decision making, resulting in 2 odds ratios (ORs) per decision characteristic.
Among the 136 respondents (109 women [80.1%] and 27 men [19.9%]; median age, 30 years [range, 18-43 years]), preferences for parent-centered decision making were positively associated with decisions that involved big-picture goals (middle OR, 2.01 [99% CI, 0.83-4.86]; high OR, 3.38 [99% CI, 1.48-7.75]) and that had the potential to harm the infant (middle OR, 1.32 [99% CI, 0.84-2.08]; high OR, 2.62 [99% CI, 1.67-4.11]). In contrast, preferences for parent-centered decision making were inversely associated with the following 4 decision characteristics: technical decisions (middle OR, 0.82 [99% CI, 0.45-1.52]; high OR, 0.48 [99% CI, 0.25-0.93]), the potential to benefit the infant (middle OR, 0.42 [99% CI, 0.16-1.05]; high OR, 0.21 [99% CI, 0.08-0.52]), requires medical expertise (middle OR, 0.48 [99% CI, 0.22-1.05]; high OR, 0.21 [99% CI, 0.10-0.48]), and a high level of urgency (middle OR, 0.47 [99% CI, 0.24-0.92]; high OR, 0.42 [99% CI, 0.22-0.83]).
Conclusions and Relevance
Preferences for parent-centered vs medical team–centered decision making among parents of infants in the NICU may vary systematically by the characteristics of particular clinical decisions. Incorporating this variation into shared decision making and endorsing models that allow parents to cede control to physicians in appropriate clinical circumstances might improve the quality and outcomes of medical decisions.
Although patients or their surrogates often want to retain control over decision making, they sometimes prefer to delegate decisions to their physicians.1-5 At least 3 classes of factors may influence a person’s preferred degree of control in medical decisions: characteristics of the decision maker, the disease, and the specific decision. Key conceptual models of the patient-physician relationship have not emphasized the influence that decision-level characteristics have on decision-making preferences.6,7 Past empirical work has explored how decision maker–related8-14 or disease-related15,16 factors influence preferences for decisional control. Whether decision-making preferences vary according to characteristics of the specific decision has received less attention.1,3,4,11 In response, Joffe and Truog17 proposed a model of medical decision making that focused on the nature of the particular decision at hand.
Little is known about how decision-level characteristics influence patients’ or surrogates’ decision-making preferences. Which characteristics of decisions influence individuals to defer to their physicians or to assume responsibility for decisions themselves is unclear. A previous qualitative study18 identified decision-level characteristics that appear to influence the preferred decision-making role among parents of infants in neonatal intensive care units (NICUs). Parents identified several characteristics that influenced their preferred parental role; 2 characteristics—a high degree of urgency and a high level of required medical expertise—were associated with a preference to delegate decisions to the medical team, whereas 4 characteristics—a high perceived risk, a high level of personal experience with the decision, involvement of foreign bodily fluids, and similarity to decisions perceived to be part of the normal parental role—were associated with a preference to retain parental control.
Patients’ desired level of control varies for different medical decisions.1,18 Clinicians and policy makers have minimal empirical evidence regarding which decisions warrant a presumption of patient (or surrogate) control and which do not. A fuller understanding of the decision-level characteristics that influence preferences is necessary to improve the quality of medical decision making. Using parents of infants in the NICU as our population of interest, we sought to identify characteristics associated with parents’ preferences for parent- vs medical team–centered decision making.
This study was conducted in the NICUs of 3 urban academic medical centers: the Hospital of the University of Pennsylvania, Pennsylvania Hospital, and the Children’s Hospital of Philadelphia. The Hospital of Pennsylvania and Pennsylvania Hospital are associated with labor and delivery units; during the study period, each had an average daily census of approximately 34 neonates. The Children’s Hospital of Philadelphia had an average daily census of approximately 85 neonates, most of whom were transferred from other NICUs. The institutional review board of the University of Pennsylvania approved the study. All participating parents provided written informed consent.
We created surveys to assess variation in decision-making preference across different types of decisions that NICU parents may face. Vignettes presented decisions about interventions including, for example, whether to perform a tracheostomy in an infant with severe chronic lung disease, whether to feed with maternal breastmilk or formula, and the choice between 2 appropriate antibiotics for neonatal sepsis. Some were urgent (emergent reintubation), whereas others were less so (placing a do-not-attempt-resuscitation order). The case vignettes were informed by a previous qualitative study18 and by the clinical experience of the neonatologist coauthors (E.M.W., N.C., and K.C.).
Each case vignette was followed by a set of ordinal-scale questions assessing decisional control preferences and parents’ judgments about the characteristics of that decision. The explanatory (independent) variables, based on the previous qualitative work,18 were those decisional characteristics hypothesized to be important in influencing parental decision-making preferences, including the degree to which a decision implicated goals of care, involved a technical means toward a goal, involved risks, had potential for benefit to the child, required medical expertise, was urgent, and was familiar based on ordinary parenting experience.
Surveys were refined using cognitive interviews with 12 participants performed by a single trained interviewer (E.M.W.) from November 7 to December 22, 2015. The findings from the cognitive interviews were incorporated into a run-in phase including 20 parents to assess for participant comprehension and consistent interpretation of study questions. Because we made no substantive changes to the instrument or methods after the run-in phase, these respondents were included in the final data set for analysis.
The final survey consisted of 16 vignettes describing a wide range of decisions that parents and physicians face in the NICU (eMethods 1 in the Supplement). The order of vignette presentation for each participant was randomized to avoid ordering effects. After each vignette, a set of 10 questions was asked (eMethods 2 in the Supplement). Two priming questions (questions 1 and 2) were followed by the dependent (question 3) and independent (questions 4-10) variables. Surveys were fielded from January 7 to July 8, 2016. Parents qualified for inclusion on their infant’s third day of life. Parents of multiple infants were eligible for inclusion; as with singleton infants, only 1 parent was eligible for survey participation, and that parent completed the survey only once. To simplify analysis and prevent within-family correlation for multiple infants, only a single value was recorded for infant-level characteristics. Most of these characteristics were necessarily identical for multiple births (ie, maternal age at delivery); for those that could differ, the value for the sicker infant was used based on preestablished criteria (eg, lower birth weight of the 2 selected) to reflect parental exposure to caring for a neonate with more severe illness. Parents were excluded from participation if they were unable to read English, were younger than 18 years, or were judged to be not capable of consent by the infant’s treating physician. Participants were offered a $20 gift card at the end of their survey session.
The dependent variable was the preferred decision maker for each decision (question 3) using the Control Preferences Scale with wording adjusted to account for parental decision makers.19 For simplicity of analysis, we dichotomized the 5-level dependent variable into parent centered (all parent, mostly parent, and equally parent and medical team) and clinician centered (mostly medical team and all medical team).
The independent variables were the responses to 7 questions about parental perceptions of characteristics of the decision (questions 4-10, answered on a scale of 1-10). We trichotomized responses into low (1-3), middle (4-7), or high (8-10) for analysis.
Other variables collected included demographic data on participants, a short set of validated health literacy questions,20 and a brief medical record review for each infant to capture clinical characteristics through the survey date. Infant characteristics included hospital, insurance coverage, prematurity level, and age in days on the survey date. Respondent (parent) characteristics included whether they had other children (and if so, whether any had been born prematurely or had chronic health problems), sex, age, educational level, marital status, ethnicity, race, and health literacy score.
To simplify sample size calculations, we assumed that responses to the independent variables would be dichotomized into equal-sized groups. We conservatively assumed that the dependent variable would be distributed unequally (ie, 20% of responses would be clinician centered). We used a significance level of .05, power of 0.90, and minimally detectable odds ratio (OR) of 2.00 in sample size calculations. We calculated the within-parent correlation using data from the run-in phase (20 parents) and assumed a conservative correlation of 0.2 in the sample size calculation. This calculation resulted in a sample size of 118 evaluable respondents. To allow for 15% missing data, we targeted a total of 136 participants.
We assessed the associations between 7 decision-level characteristics and parents’ preferences for parent-centered vs medical team–centered decision making, accounting for clustering of vignettes within respondents. We used logistic regression models to determine which characteristics were associated with the dependent variable. Characteristics under consideration included all 7 decisional characteristics and 12 infant and respondent characteristics. For the dependent variable, we arrived at the final model in 3 steps. In step 1, we ran a model including the 7 decisional characteristics. In step 2, we added each infant and respondent characteristic separately to the model, 1 variable at a time (ie, each step 2 model included the 7 decisional characteristics and 1 infant or respondent characteristic). In step 3, those infant and respondent characteristics with P < .20 (using the χ2 test) in step 2 were added to the model; variables were then removed sequentially using backward selection until all remaining variables had P < .05. This process gave the final model. We report 99% CIs for the ORs to account for multiple testing. Because responses to the decision-maker preference questions were likely to be correlated within parent respondents, we used generalized estimating equations to examine and compare the unadjusted and adjusted associations between the independent and dependent variables. Analysis was performed using SAS software (version 9.3; SAS Institute Inc).
Of 209 parents who underwent screening, 200 eligible parents were approached from January 7 to July 8, 2016; 136 of these parents (68.0%) were enrolled and included in the analysis (109 women [80.1%] and 27 men [19.9%]; median age, 30 years [range, 18-43 years]). Participants were diverse with respect to Medicaid insurance coverage, self-reported race, income, marital status, and educational level. Infants represented a wide range of gestational ages (23-41 weeks). Table 1 shows characteristics of the infants and of their parent respondents.
Parents’ decision-making preferences varied across the 16 vignettes (Figure). Mean responses along the 5-point Control Preferences Scale ranged from mostly parent to all medical team. On average, parents preferred to retain control for 7 decisions (research participation, do-not-attempt-resuscitation orders, deciding between breastmilk and formula, withdrawal of life-sustaining treatment, choice between medications with markedly different adverse effect profiles, tracheostomy, and use of a home apnea monitor). In contrast, parents generally preferred to delegate 9 decisions to their medical teams (blood transfusion for apnea, placement of a peripherally inserted central catheter, lumbar puncture, surgery for esophageal atresia, elective intubation, umbilical line insertion, choice between similar inotropes, choice between similar antibiotics, and emergent intubation).
Two characteristics of decisions, including those involving big-picture goals (middle OR, 2.01 [99% CI, 0.83-4.86]; high OR, 3.38 [99% CI, 1.48-7.75]) and with the potential to harm the infant (middle OR, 1.32 [99% CI, 0.84-2.08]; high OR, 2.62 [99% CI, 1.67-4.11]), were associated with a preference for parent-centered decision making. In contrast, 4 characteristics, including technical decisions (middle OR, 0.82 [99% CI, 0.45-1.52]; high OR. 0.48 [99% CI, 0.25-0.93]), decisions with the potential to benefit the infant (middle OR, 0.42 [99% CI, 0.16-1.05]; high OR, 0.21 [99% CI, 0.08-0.52]), decisions requiring medical expertise (middle OR, 0.48 [99% CI, 0.22-1.05]; high OR, 0.21 [99% CI, 0.10-0.48]), and decisions with a high level of urgency (middle OR, 0.47 [99% CI, 0.24-0.92]; high OR, 0.42 [99% CI, 0.22-0.83]), were associated with a preference for medical team–centered decision making (Table 2). Prior parental experience with the decision (middle OR, 1.05 [99% CI, 0.67-1.63]; high OR, 0.87 [99% CI, 0.56-1.37]) was the only decision characteristic that was not significantly associated with a preferred decision-making role in the multivariable analysis.
Among the 12 infant and parent-respondent characteristics considered for inclusion in the multivariable model, only having other children was marginally associated (χ22 = 9.08; P < .05) with a preference for parent-centered decision making. None was associated with a preference for medical team–centered decision making after adjusting for other variables.
We asked whether decision-level characteristics are associated with preferences for decision-making control among parents of infants receiving neonatal intensive care. Consistent with our a priori hypotheses, we found that characteristics of specific medical decisions were associated with preferences for parent-centered vs medical team–centered decision making across a wide range of decisions relevant to NICU parents. These findings have important implications for shared decision making. Research and clinical evaluation of decision-making preferences must consider decision-level characteristics and cannot assume that parents want similar control over a diverse range of decisions.
Parents chose to delegate decision-making control to the medical team for 9 of the 16 decisions presented to them. This preference to delegate more than half of decisions to the medical team was not anticipated. However, these findings are consistent with early empirical studies that found that adult patients wanted less decision-making control than their physicians believed they did.3 Further work must explore how clinicians and policy makers should operationalize this finding in the NICU, for example, when selecting which interventions warrant written consent, verbal agreement, informal discussion, or no parental involvement.
Decisions that parents viewed as involving big-picture goals were associated with a preference for parent-centered decision making, whereas those that they viewed as being technical in nature were associated with a preference for medical team–centered decision making. This finding is consistent with the conceptual model of Joffe and Truog,17 which emphasizes the distinction between ends-oriented and means-oriented decisions. These findings further suggest that shared decision making should focus in particular on decisions that implicate goals of care.
Decisions believed by parents to have the potential to cause harm were associated with a preference for parent-centered decision making. Risk has long been assumed to be an important contributor to preferences for decision-making control in a range of medical scenarios.21 Consent for medical interventions often aims to be more comprehensive and formal as the risk increases. Our findings suggest that, consistent with such practices, parents of infants in the NICU believe that risky interventions warrant greater parental involvement in the decision-making process.
Our data also suggest the converse: decisions believed by parents to be most likely to benefit the infant were associated with a preference for clinician-centered decision making. Conventional approaches to shared decision making often do not take these considerations into account. In neonatology, many common, necessary, low-risk, and high-benefit interventions require parental involvement through consent. For example, in extremely premature neonates, peripherally inserted central catheters are a high-benefit, necessary, and relatively low-risk intervention but often require formal parental consent. In contrast, the use of postnatal corticosteroids in the same population is of uncertain benefit and has substantial neurodevelopmental risk. Although the most recent American Academy of Pediatrics policy statement on this decision recommends a shared decision-making approach,22 a single-center study23 found low rates of informed consent documentation for the use of postnatal corticosteroids for premature infants. At present, in the case of peripherally inserted central catheters and postnatal corticosteroids, consent requirements are based on historical precedent and legal concerns. If the risk-benefit level is as important as our findings suggest, neonatology and medicine more broadly may benefit from a reevaluation of which interventions warrant patient or surrogate decision-making involvement and formal consent.
Decisions that parents viewed as urgent were associated with a preference for clinician-centered decision making. This finding is consistent with usual practice in emergency care, in which usual communication or consent processes may be suspended or altered when the intervention is urgent. Our data suggest that parents accept the need to forgo or limit detailed formal consent in such situations.
We considered infant and parent-respondent characteristics for inclusion in the multivariable model. The finding that having a previous child, especially one in the NICU or with a chronic illness, was marginally associated with a preference for parent-centered decision making recalls the observation that parents of former premature neonates prefer more aggressive care compared with parents of former term infants.24 This association may be owing to the knowledge or confidence that experienced parents, compared with new parents, bring to the medical context. Further empirical work is needed to confirm this finding and to evaluate how prior experience with child rearing in general and with caring for a medically complex child in particular may influence parents’ decision-making preferences for their other children. Of importance, no other measured infant or parent characteristic was associated with preferences for parent- or medical team–centered decision making.
Our findings confirm that parents often want to delegate decision-making power. Additional research is needed to assess whether current approaches result in physicians imposing unwanted decisions on patients25 and their surrogates and how this occurrence may relate to decisional burden.26-29 Our data also challenge the claim that “the types of decisions that call for different approaches cannot be categorized because each patient is different and it is the patient, not the decision under consideration, that guides the process.”7(p904) More broadly, our findings raise fundamental questions about whether consent practices in neonatology and likely other areas of pediatrics and medicine serve primarily to protect clinicians and institutions or to empower parents to make decisions for their infants in accordance with their values and priorities. Recent work considering a broadened understanding of the functions of consent or new consent models has begun to explore these issues.30-32
The study reported herein has several limitations. First, it was administered in a single urban area, albeit at 3 distinct institutions with diverse patient populations. Second, the survey design did not allow for consideration of other decision-level characteristics that may be important to some parents in influencing their decision-making preferences. Third, although high proportions of non-Hispanic black and non-Hispanic white parents participated, other groups were not well represented, potentially limiting generalizability. However, respondents represented a wide range of household income, educational level, and marital status. Fourth, although the rate of nonparticipation was low, those who did not participate might have differed meaningfully from participants. Finally, preferences may differ in other decision-making situations, such as with parents of older children or adults making their own medical decisions. Whether similar associations between decisional characteristics and preferred approaches to decision making are present in other populations of medical decision makers requires further research.33,34
Characteristics of particular decisions were associated with preferences for parent-centered vs medical team–centered decision making among NICU parents across a wide variety of clinical scenarios. These findings suggest that, within neonatal intensive care, new models of medical decision making that align with parental preferences must be considered. Future work should evaluate the potential benefit of incorporating these preferences into clinical care, for example, when selecting whether interventions warrant written consent, verbal agreement, informal discussion, or no parental involvement. Future work must also assess the association between decisional preferences and decision characteristics in other populations of medical decision makers.
Accepted for Publication: December 21, 2017.
Corresponding Author: Elliott Mark Weiss, MD, MSME, Division of Neonatology, Department of Pediatrics, University of Washington School of Medicine, 4800 Sand Point Way NE, Mail Stop FA.2.113 Neonatology, Seattle, WA 98105 (firstname.lastname@example.org).
Published Online: March 19, 2018. doi:10.1001/jamapediatrics.2017.5776
Author Contributions: Dr Weiss had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Weiss, Cook, Joffe.
Acquisition, analysis, or interpretation of data: Weiss, Xie, Coughlin, Joffe.
Drafting of the manuscript: Weiss.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Weiss, Xie.
Administrative, technical, or material support: Weiss, Coughlin, Joffe.
Study supervision: Weiss, Cook, Joffe.
Conflict of Interest Disclosures: None reported.
Additional Contributions: We thank the parents for their participation in this study and the neonatal intensive care unit physicians and nurses for facilitating access to their families. Katharine Gleason, MPH, University of Pennsylvania, provided support with data analysis and editing and was not compensated. Lorine Vogt, BSN, University of Pennsylvania, assisted with recruitment and survey administration, for which she received nominal compensation. Frances K. Barg, PhD, and the staff at the University of Pennsylvania Mixed Methods Research Laboratory provided support with survey development and cognitive interview administration and were not compensated. Aspen Miller provided support with literature review and was not compensated.