Quality of life (QoL) data give insight into the patient perspective of disease impact, helping to align patient and provider treatment goals. Despite their extensive use in clinical trials, QoL measures are rarely used in clinical management of asthma. One reason may be the lack of clarity on how to interpret scores and use this data to enhance clinical decision-making. The Patient-Reported Outcomes Measurement Information System® (PROMIS®) includes a set of measures for assessing a range of physical, mental and social QoL domains in adults and children that were developed using rigorous qualitative and quantitative methodology and are available for use by clinicians and researchers (1). PROMIS Asthma Impact is a well-validated instrument (2, 3) that assesses asthma-related QoL over the prior seven days, with 5-point Likert scale responses to items such as, “I was bothered by the amount of time I spent wheezing.” Unlike other commonly used control measures like Global Initiative for Asthma (GINA) control criteria (4) and the Asthma Control Test (ACT)(5), which focus primarily on asthma-related impairment, PROMIS Asthma Impact provides a fuller perspective of how asthma impacts patients’ lives and may complement other control measures. However, there are a lack of established cut points for interpreting the scores.
To address this limitation, we performed a secondary analysis of longitudinal data from an observational study of children 8 to 17 years of age with asthma (6) to assess the relationship between PROMIS Asthma Impact and commonly used measures of asthma control to establish relevant cut points. Global Initiative for Asthma (GINA) control criteria (4) and the Asthma Control Test (ACT) and childhood Asthma Control Test (cACT) (5) are widely used both clinically and in research to assess symptoms, rescue medication use and activity limitation over the prior four weeks. Using the GINA assessment to define control, the parent study recruited children with uncontrolled or partly controlled asthma (6)to examine the association between physical activity (using pedometry) and asthma control as previously described (6). GINA assessment and ACT (or cACT) were performed at the initial study visit and were repeated at the final visit 28 days later. PROMIS Asthma Impact was administered via online survey on days 7, 14, 21, and 28.
We assessed correlations between PROMIS Asthma Impact scores (days 7 and 28), ACT/cACT (days 1 and 28) and GINA assessments (days 1 and 28) using Pearson product moment and polyserial correlation coefficients depending on the scale of the scores. We used Receiver Operating Characteristics (ROC) curves to identify optimal cut points of PROMIS Asthma Impact T-Score for detecting uncontrolled asthma. Our anchor was dichotomous classification of ACT/cACT - score ≤ 19 for uncontrolled versus 20+ for well-controlled, or trichotomous classification of GINA: 1) uncontrolled, 2) partly controlled and 3) well controlled. Generalized mixed logistic regression models were used to estimate probability of control classification of each participant predicted by PROMIS Asthma Impact T-score. A random intercept was included in each model to account for the data within-subject dependency among the repeated assessments. ROCs were constructed using the probability estimates computed from the mixed models. Area Under Curve (AUC) was used to quantify how well the PROMIS Asthma Impact T-score could distinguish between the classes. For the GINA, pairwise AUCs and accumulative AUC (the average of the pairwise AUCs) were estimated (7). Two methods or criteria were applied for determining the optimal cut points: 1) The distances to the top-left corner from points on the ROC curve were calculated, and the point with the lowest distance was chosen for the cut-off point; 2) Youden Indices were calculated, and the point with the highest Youden Index was chosen (8). We used SAS version 9.4 (SAS Institute, Cary, NC) and a 2-tailed significance level of α ≤ 0.05 for all assessments.
The parent study enrolled 105 participants, with a mean age of 11.4 (SD 2.6) years, 53% male, 45% Black race and 17% Hispanic/Latinx ethnicity. At the first study visit, 12 (11%) participants had well controlled asthma, 57 (54%) as partly controlled and 36 (34%) as uncontrolled. Moderate to strong negative correlations were observed between PROMIS Asthma Impact scores and ACT/cACT scores (ACT: r = −0.53, 95% confidence interval (CI): (− 0.72, −0.25) day 0/7, r=−0.84, 95% CI (−0.91, −0.72) day 28; cACT: r = −0.72, 95% CI (−0.83, − 0.55) day 0/7, r=−0.82, 95% CI (−0.89, −0.68) day 28), with lower ACT/cACT scores and higher PROMIS Asthma Impact scores indicating poorer asthma control. A moderate negative correlation was also observed between PROMIS Asthma Impact scores and GINA assessment, such that as GINA score went up (indicating better asthma control), PROMIS Asthma Impact scores decreased, indicating better asthma control (r = −0.61, 95% CI (−0.76, −0.46) day 0/7, r = − 0.68, 95% CI (−0.81, −0.54) day 28. The ROC/AUC analyses (Table 1) indicated a PROMIS Asthma Impact T-score of 45.6 as the optimal cut point to distinguish uncontrolled from well controlled asthma using ACT/cACT scores. Using the GINA assessment, a PROMIS Asthma Impact T-score of 38.7 would be the optimal cut point to distinguish partly controlled from well-controlled asthma and a T-score of 49.2 to distinguish partly controlled from uncontrolled asthma (Figure 1).
Table 1.
Calculated PROMIS Asthma Impact T-Score Cut-Points
| Control Classifier | Control Category | AUC | PROMIS Asthma Impact T-Score | Sensitivity | Specificity | |
|---|---|---|---|---|---|---|
| ACT/cACT | Uncontrolled | Well controlled | 0.950 | 45.6 | 0.932 | 0.850 |
| GINA | Partly controlled | Uncontrolled | 0.709 | 49.2 | 0.827 | 0.843 |
| Partly controlled | Well controlled | 0.661 | 38.7 | 0.708 | 0.727 | |
| Uncontrolled | Well controlled | 0.909 | ||||
| Multi-level Accumulative | 0.760 | |||||
Figure 1.
Suggested cut points in PROMIS Asthma Impact T-Score for defining asthma control using (A) the Asthma Control Test (ACT) or Childhood Asthma Control Test (cACT) and (B) the Global Initiative for Asthma (GINA) asthma control criteria.
In a diverse sample of children and adolescents, an asthma-specific QoL measure showed a moderate to strong correlation with GINA and with ACT/cACT score-derived classification of asthma control. However, a substantial portion of the variance between PROMIS Asthma Impact and ACT score or GINA classification was unexplained, suggesting that the QoL measure provides distinct information about asthma control that is not captured by other asthma control measures. We generated PROMIS Asthma Impact T-score cut points for identifying suboptimal asthma control that can aid clinicians in interpretation of this QoL measure. Generalized mixed models allowed us to incorporate multiple assessments from each participant to inform our estimation of the threshold values in PROMIS Asthma Impact, a stronger approach than using a single time point of data.
Our study has several limitations. In addition to small sample size and short duration of the study period, our sample was likely biased towards children with more severe, poorly controlled asthma due to inclusion criteria of the parent study. Future studies should include a larger number of well controlled asthmatics to better validate score cut points in this population. Similarly, larger confirmatory studies would allow for examination of potential sub-group differences in cut point scores. As a secondary analysis of data, our comparisons of PROMIS Asthma Impact with GINA assessment and ACT/cACT may be affected by the different reference periods used (7 days vs 4 weeks, respectively). However, respondents are typically impacted in their ratings by a recency effect, where the most recent experiences tend to have the greatest influence on responses(9). Finally, commonly used asthma control measures like the ACT/cACT and GINA assessment are weighted heavily towards the impairment domain of asthma control without addressing the risk domain, highlighting the need for more comprehensive control assessment metrics. While PROMIS Asthma Impact is also predominantly focused on impairment, it includes items that assess the occurance of “asthma attacks” and missed school days because of asthma, elements of the risk assessment, a unique strength among the three asthma control measures described herein.
In conclusion, the assessment of asthma control is complex, and our findings support the use of a multicomponent assessment of control in children and adolescents. Asthma QoL measures like PROMIS Asthma Impact may offer additional information and insight to asthma control that complement the GINA and ACT/cACT assessments by incorporating patient perspectives.
Clinical Implications:
Quality of life (QoL) measures incorporate patient perspectives that complement traditional metrics used to detect uncontrolled asthma. We identified optimal cut points in PROMIS® Asthma Impact score for identifying uncontrolled asthma in a diverse sample of children with asthma.
Funding Acknowledgements:
Research reported in this publication was supported by NIAMS (U19AR069525), NHLBI R01HL135235, and NCATS UL1TR002489 and KL2TR002490. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. The parent study made use of systems and services provided by the Patient-Reported Outcomes Core (PRO Core; pro.unc.edu) at the Lineberger Comprehensive Cancer Center (LCCC) of the University of North Carolina. PRO Core is funded in part by a National Cancer Institute Cancer Center Core Support Grant (5-P30-CA016086) and the University Cancer Research Fund of North Carolina. The LCCC Bioinformatics Core provides the computational infrastructure for the PRO Core system.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Disclosure Statement: M.L.H. reports receiving consulting fees from GlaxoSmithKline (GSK).A.J.B., N.L., L.L., and B.B.R. report no conflicts.
REFERENCES
- 1.Reeve BB, Hays RD, Bjorner JB, Cook KF, Crane PK, Teresi JA, et al. Psychometric evaluation and calibration of health-related quality of life item banks: plans for the Patient-Reported Outcomes Measurement Information System (PROMIS). Med Care. 2007;45(5 Suppl 1):S22–31. [DOI] [PubMed] [Google Scholar]
- 2.Howell CR, Thompson LA, Gross HE, Reeve BB, DeWalt DA, Huang IC. Responsiveness to Change in PROMIS((R)) Measures among Children with Asthma: A Report from the PROMIS((R)) Pediatric Asthma Study. Value Health. 2016;19(2):192–201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Nelson A, Singh A, Dasgupta M, Simpson PM, Chiu A, Brousseau DC, et al. Detection of changes of functioning over time after asthma exacerbation in children with the use of PROMIS domains. J Asthma. 2022;59(10):1981–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Global Initiative for Asthma (GINA). Global Strategy for Asthma Management and Prevention. 2019. Accessed 12 December 2022. Available from: https://ginasthma.org.
- 5.Schatz M, Sorkness CA, Li JT, Marcus P, Murray JJ, Nathan RA, et al. Asthma Control Test: reliability, validity, and responsiveness in patients not previously followed by asthma specialists. J Allergy Clin Immunol. 2006;117(3):549–56. [DOI] [PubMed] [Google Scholar]
- 6.Hernandez ML, Lucas N, Mann C, Lin L, Burbank AJ, Brown J, et al. Association of step count with PROMIS pediatric health-related quality of life measures in children and adolescents with persistent asthma. J Allergy Clin Immunol Pract. 2021;9(6):2492–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hand DJ, Till RJ A simple generalisation of the area under the ROC curve for multiple class classification problems. Machine Learning. 2001;45(2):171–86. [Google Scholar]
- 8.Hajian-Tilaki K Receiver Operating Characteristic (ROC) curve analysis for medical diagnostic test evaluation. Caspian J Intern Med. 2013;4(2):627–35. [PMC free article] [PubMed] [Google Scholar]
- 9.McClendon MJ. Response-order effects for dichotomous questions. Social Science Quarterly. 1986;67:205–11. [Google Scholar]