Evidence-Based Infant Assessment for Cerebral Palsy: Diagnosis Timelines and Intervention Access in a Newborn Follow-up Setting

Abstract

Evidence-based assessment pathways inform early detection of cerebral palsy and access to intervention. This study investigated the relationships between early evidence-based assessments, diagnosis timeline, and rehabilitation intervention access in a population of children with cerebral palsy who were seen between 2010 and 2022 at the University of Wisconsin Waisman Center Newborn Follow Up Clinic. Cerebral palsy–specific assessments were increasingly integrated after the publication of early detection guidelines by Novak et al. in 2017. Age at cerebral palsy first mention (high risk for cerebral palsy) decreased over time, although age at diagnosis remained similar. Infants who received multiple evidence-based assessments were diagnosed at a younger age. Ninety-nine percent of children were referred to rehabilitation therapies before diagnosis. Infant age at referral to outpatient therapies decreased over time. This study provides novel clinical data on diagnosis timelines and identifies remaining gaps related to implementation feasibility toward improved early diagnosis and intervention access.

Recent clinical guidelines highlight the predictive validity and clinical importance of a set of key evidence-based assessments for early detection of cerebral palsy, with the possibility of accurately diagnosing the condition within 6 months’ corrected age.¹ Early diagnosis of cerebral palsy has several important benefits. Parents and caregivers prefer earlier diagnosis,² and diagnosis promotes parent engagement in therapies and bonding.^3,4 An early diagnosis can provide access to resources including early cerebral palsy–specific rehabilitation interventions,¹ financial assistance, and other services such as parent and caregiver mental health support.^1,5 Emerging evidence surrounding early targeted interventions indicates the potential for improved outcomes and reduced comorbidities.^5,6

For infants with newborn-detectable risks (e.g., infants with known risk factors for cerebral palsy at age <5 months), a standardized assessment pathway can be implemented within the first few months of life.¹ Key evidence-based assessments include magnetic resonance imaging (MRI), the General Movements Assessment, and the Hammersmith Infant Neurological Examination. MRI can identify injury patterns associated with an increased risk of cerebral palsy, such as white matter injury, cortical or deep gray matter lesions, and brain maldevelopment.^1,7 The General Movements Assessment evaluates spontaneous infant movement patterns via video recording and can be performed from the preterm period up to 5 months postterm.⁸ In the “Fidgety” movement period (around 9–20 weeks postterm age), the General Movements Assessment has high predictive value, with approximately 98% sensitivity and 91% specificity for cerebral palsy.⁷ The Hammersmith Infant Neurological Examination is a standardized neurologic assessment that assesses cranial nerve function, posture, movement, muscle tone, and reflexes in infants 3–24 months’ corrected age. Hammersmith Infant Neurological Examination optimality scores are determined by age, and established cutoff scores are 90% predictive of cerebral palsy.⁹ A combination of these assessments (MRI, General Movements Assessment, Hammersmith Infant Neurological Examination) provides greater accuracy in diagnosis.¹ For infants not referred for assessment until after 5 months of age, MRI and the Hammersmith Infant Neurological Examination remain most strongly predictive of cerebral palsy. According to Novak et al., motor assessment tools, including the Developmental Assessment of Young Children, the Alberta Infant Motor Scale, or the Neuro-Sensory Motor Developmental Assessment, may also assist in predicting outcomes in infants who are referred for assessment later.¹

Studies evaluating the clinical implementation of the Novak et al. early detection guidelines have reported success in lowering the age at diagnosis with the use of varied implementation strategies.^10–12 Many of these studies have been conducted in a network of high-risk infant follow-up clinics in the United States,¹⁰ although recent publications have highlighted the feasibility of lowering the age of diagnosis in other clinical settings as well, including internationally.^13–16 Studies of guideline implementation in varied clinical settings allow further analysis of the feasibility of implementing evidence-based assessment tools in diverse practice settings and patient populations, and subsequent impact on diagnosis timelines and outcomes. This retrospective study aimed to determine the relationships between early evidence-based assessment, diagnosis timeline, and intervention access in a population of infants who attended a newborn follow-up clinic in Madison, Wisconsin, and were diagnosed with cerebral palsy.

Patients and Methods

Eligible patients were determined by a medical record search for patients with diagnosis codes relating to cerebral palsy who had attended at least 1 visit to the University of Wisconsin Waisman Center Newborn Follow-Up Clinic (NBFU Clinic), an interdisciplinary outpatient clinic specializing in assessment for children aged 0–3 years at high risk for developmental disabilities. Following publication of the 2017 guidelines, the NBFU Clinic team made a concerted effort to increase use of evidence-based assessments and support early diagnosis. The clinic provided funds and time for providers to be trained in the General Movements Assessment throughout 2017, and by 2018 providers were routinely using the General Movements Assessment during clinic visits. Efforts were made to schedule clinic visits during the age range for the General Movements Assessment (either as an interdisciplinary visit, a physician-only visit, or as a therapy assessment through partnering rehabilitation clinics with training in the General Movements Assessment).

To conduct the retrospective medical record review, records were searched using International Classification of Diseases, Tenth Revision (ICD-10) codes, the search term “cerebral palsy,” and a location code of developmental pediatrics, the primary provider at the NBFU Clinic. Patients were included if they had a cerebral palsy diagnosis documented in their medical record between 2010 and 2022, attended at least 1 appointment at the NBFU Clinic, and if the date and provider of their cerebral palsy diagnosis were explicitly documented in their medical record. Patients were included if their diagnosis was provided by a developmental pediatrician at the NBFU Clinic, or by another medical provider within the University of Wisconsin health system (UW-Health). Patients were excluded if the cause of their cerebral palsy was attributed to an event outside of the perinatal period (eg, a traumatic brain injury or surgery), or if cerebral palsy was discussed but not formally diagnosed. This study was reviewed by the Institutional Review Board at the University of Wisconsin–Madison and granted exempt status.

Medical record review was led by authors with clinical backgrounds (ES, JL, and MV) to identify (1) corrected age at first documented conversation indicating high risk for cerebral palsy (age at cerebral palsy first mention); (2) corrected age at cerebral palsy diagnosis; (3) provider type; (4) assessments completed prior to diagnosis, with a specific review for MRI, General Movements Assessment, and Hammersmith Infant Neurological Examination; (5) referral dates for rehabilitation therapies (early intervention and outpatient therapies); (6) Gross Motor Function Classification System (GMFCS) level; and (7) cerebral palsy distribution and subtype. If Gross Motor Function Classification System scores were unavailable and the child was at least 2 years old at the time of medical record review, physician and/or physical therapist study team members provided a Gross Motor Function Classification System rating based on available medical record information.

Analyses of intervention referral dates were completed for patients for whom a dated referral was documented in the medical record, regardless of whether the patient proceeded with scheduling or initiating therapies. If it was noted that a patient was participating in therapies at the time of their diagnosis but a referral date was not available, the patient was included in descriptive statistics summarizing services received, but excluded from analyses of age at referral. Referrals to “outpatient therapy” included referrals for physical, occupational, and/or speech therapy at a clinic-based location. Referrals noted as “early intervention” included referrals to the Wisconsin Birth to 3 program or comparable early intervention program in neighboring states. Analyses were also performed with first referral date, indicating the date of earliest intervention referral regardless of service type.

Data were analyzed with SPSS version 28 and R Studio version 2023.03.0. Participant characteristics, assessment data, and therapies received were summarized with descriptive statistics. All statistical analyses were assessed for significance at α = 0.05. All ages were corrected for prematurity for children born at <37 weeks and are presented in months. Ages at cerebral palsy first mention, cerebral palsy diagnosis, and referrals to intervention were compared using unpaired t tests for participants who were diagnosed between 2010 and 2017 and between 2018 and 2022. Linear regression was performed to further evaluate the relationship between diagnosis year and age at referral. Patients were also grouped by the number of evidence-based assessments received among 3 key assessments: MRI, General Movements Assessment, and Hammersmith Infant Neurological Examination (0, 1, or 2 + tests). Ages at cerebral palsy diagnosis, cerebral palsy first mention, and referrals to intervention were compared among groups with a 1-way analysis of variance. To consider the impact of receiving additional guideline-recommended motor assessments when the General Movements Assessment and Hammersmith Infant Neurological Examination were unavailable, patients were grouped by (1) those who received MRI + General Movements Assessment and/or Hammersmith Infant Neurological Examination; (2) those who received MRI + Alberta Infant Motor Scale, Test of Infant Motor Performance, and/or Developmental Assessment of Young Children (not General Movements Assessment or Hammersmith Infant Neurological Examination); (3) those who received 1 or fewer assessments. These assessments (Alberta Infant Motor Scale, Test of Infant Performance, Developmental Assessment of Young Children) were selected because they were the additional guideline-recommended assessments performed in the sample. Ages at cerebral palsy diagnosis and cerebral palsy first mention were compared with a 1-way analysis of variance. Additionally, linear regression analysis was performed to evaluate the relationship between the number of evidence-based assessments received (0, 1, 2 + ) and age at cerebral palsy first mention or cerebral palsy diagnosis after correcting for sex, race, and Gross Motor Function Classification System level. Patients with missing data for 1 or more predictors were excluded from regression analysis.

Finally, a series of exploratory linear regression analyses were conducted to identify independent predictors of age at cerebral palsy first mention and age at diagnosis. Because the distributions of age at cerebral palsy first mention and age at diagnosis were skewed, a log-transformation was applied before conducting the regression analyses. Because age at cerebral palsy first mention included negative values (in some infants born preterm), a constant was added to all data such that the minimum value was 0.001 before log transforming the data. Both forward and backward regression variable selection procedures with a P value selection criterion of P <.10 were used to determine parsimonious models. If the final parsimonious models derived from the forward and backward selection procedures did not concur, the model with the highest R² (best fitting model) value was selected as the final model. The following candidate predictors were included in the initial nonparsimonious regression model: (1) number of evidence-based assessments among MRI, General Movements Assessment, Hammersmith Infant Neurological Examination (0, 1, 2 + ), (2) year of diagnosis (separated into six bins), 3) cerebral palsy distribution (hemiparesis, diparesis, tri/quadriparesis), (4) cerebral palsy subtype (spastic, dyskinetic, mixed/other), (5) Gross Motor Function Classification System level, (6) Preterm (<37 weeks)—yes/no, (7) corrected age at first visit to the NBFU Clinic, (8) known pathology on MRI—yes/no, (9) patient’s race, as a social construct potentially relating to access to care, and (10) diagnosing provider (developmental pediatrician, pediatric neurologist, physical medicine and rehabilitation [PM&R], other). For the model, cerebral palsy distribution was included for dyskinetic and mixed cerebral palsy based on affected body regions as described in the medical record. Patients with missing data for 1 or more predictors were excluded. Model assumptions of the final model were verified by examining residual plots.

Results

Ninety-one patients were included in the final data set. Patients spanned all Gross Motor Function Classification System levels and had varied cerebral palsy distributions and subtypes (Table 1). Patients were divided into 2 cohorts (2010–2017, 2018–2022) based on the year of their cerebral palsy diagnosis. MRI, General Movements Assessment, and Hammersmith Infant Neurological Examination were integrated more frequently in the later cohort (Figure 1).

Table 1.

Participant Characteristics (n = 91).

	n	%
Sex
Male	46	50
Female	45	50
Race
Black	12	13
Asian	2	2
White	71	78
Other	4	4
Not reported	2	2
Ethnicity
Hispanic or Latinx	10	11
Not Hispanic or Latinx	78	86
Not Reported	3	3
CP distribution (spastic or mixed spastic-dyskinetic)
Unilateral: hemiparesis	27	35
Bilateral: diparesis	14	18
Bilateral: quadriparesis	33	43
Other	2	2
Unknown	3	3
CP Subtype
Spastic	67	74
Dyskinetic	5	5
Mixed	15	16
Other	2	2
Unknown	2	2
GMFCS
1	25	28
2	14	15
3	14	15
4	13	14
5	17	19
Unknown/too young	8	9
Diagnosing provider
Developmental pediatrician	69	76
Neurologist	13	14
PM&R	5	5
Other	4	4

Abbreviations: CP, cerebral palsy; GMFCS, Gross Motor Function Classification System; PM&R, physical medicine and rehabilitation.

Figure 1.

Assessments used before/after publication of early diagnosis guidelines (Novak 2017). 2010–2017 subgroup n = 44, 2018–2022 subgroup n = 47. AIMS, Alberta Infant Motor Scale; DAYC, Developmental Assessment of Young Children; GMA, General Movements Assessment; HINE, Hammersmith Infant Neurological Examination; MRI, Magnetic Resonance Imaging; TIMP, Test of Infant Motor Performance.

The average age of cerebral palsy diagnosis was comparable between the earlier and later cohorts (2010–2017: 14.7±7.1 months, 2018–2022: 15.5±8.1 months). The age at cerebral palsy first mention was lower in the later cohort (2010–2017: 11.0±8.7, 2018–2022: 6.3±6.1 months). Across all years, there was a difference in age at diagnosis among children who received none (22.1±8.5 months), 1 (15.5±7.6 months), or 2 or more (12.7±6.1 months) of the 3 key evidence-based assessments (MRI, General Movements Assessment, Hammersmith Infant Neurological Examination). There was also a difference in age at cerebral palsy first mention among children who received none (16.3±7.3 months), 1 (9.9±8.3 months), or 2 or more (4.5±4.1 months) of the 3 evidence-based assessments (Figure 2). Number of evidence-based assessments (MRI, General Movements Assessment, Hammersmith Infant Neurological Examination) was a significant predictor of age at cerebral palsy first mention and age at diagnosis after adjusting for sex, race, and Gross Motor Function Classification System level (P < .05, n = 82). When other motor assessments (Alberta Infant Motor Scale, Test of Infant Motor Performance, and Developmental Assessment of Young Children) were considered, the group of children who received MRI + General Movements Assessment and/or Hammersmith Infant Neurological Examination and the group of children who received MRI + Alberta Infant Motor Scale, Test of Infant Motor Performance, and/or Developmental Assessment of Young Children were significantly younger at age of cerebral palsy first mention compared to children who received 1 or fewer evidence-based assessments (P < .001). Difference in age at diagnosis approached significance among these groups (P = .07, Supplemental Figure 1).

Figure 2.

Age at cerebral palsy first mention and diagnosis. (A) Age at cerebral palsy first mention grouped by year of diagnosis; (B) age at diagnosis grouped by year of diagnosis; (C) age at cerebral palsy first mention grouped by number of evidence-based assessments received (of MRI, GMA, HINE); (D) age at diagnosis grouped by number of evidence-based assessments received (of MRI, GMA, HINE). *t test P < .05, ^†1-way analysis of variance, Tukey post hoc test, P <.05. GMA, General Movements Assessment; HINE, Hammersmith Infant Neurological Examination; MRI, Magnetic Resonance Imaging.

Exploratory regression analyses were run on a sample of N = 80 after removing patients with missing data for 1 or more predictors. The best-fit regression model for age at cerebral palsy first mention included the following significant predictors: corrected age at first visit to the NBFU Clinic, pathology indicated on MRI, cerebral palsy distribution, and preterm birth (R² = 0.259; Table 2). The best-fit regression model for age at diagnosis of cerebral palsy included number of tests among MRI, General Movements Assessment, Hammersmith Infant Neurological Examination, year of diagnosis, diagnosing provider, and pathology indicated on MRI as significant predictors (R² = 0.297; Table 2).

Table 2.

Best-fit Models From Exploratory Regression Analysis (n = 80).^a

	Estimate	SE	P value
CP First Mention
Intercept	3.03	0.47	<.001*
Preterm (yes)	0.71	0.27	.007*
Age at first visit	0.08	0.03	.002*
Pathology on MRI (yes)	−1.04	0.38	.006*
CP distribution (reference: hemiparesis)			.015*
Diparesis	−1.10	0.38
Tri- or quadriparesis	−0.23	0.29
CP diagnosis
Intercept	3.09	0.19	<.001*
Number of assessments	−0.34	0.11	.003*
Year of diagnosis (reference:  2010–2012)			.008*
2013–2014	0.01	0.20
2015–2016	−0.06	0.20
2017–2018	0.21	0.20
2019–2020	0.61	0.22
2021–2022	0.23	0.21
Diagnosing Provider (reference: developmental pediatrician)			.001*
Neurologist	−0.34	0.14
Physical medicine & rehabilitation	0.12	0.23
Other	−0.80	0.23
Pathology on MRI (yes)	−0.25	0.15	.099*

Abbreviations: CP, cerebral palsy; GMFCS, Gross Motor Function Classification System; MRI, magnetic resonance imaging; SE, standard error.

^aOutcomes (CP first mention, CP diagnosis) were log transformed.

^*All P < .1.

Ninety-nine percent of children were referred to rehabilitation therapies before or on their date of diagnosis. Eighty-seven percent were referred to early intervention and 77% were referred to at least 1 outpatient therapy service (physical therapy, occupational therapy, or speech and language pathology) before or on their diagnosis date (Table 3). The average age of first referral to a rehabilitation service (early intervention or outpatient) was 2.59 ± 3.39 months (n = 75). Adjusted age at referral to outpatient therapies was negatively associated with diagnosis year (r = −0.435, P < .001, n = 66), whereas referral to early intervention was positively associated with diagnosis year (r = 0.330, P = .009, n = 62) (Figure 3). Age at outpatient referral significantly differed among those receiving none, 1, or 2 or more of the evidence-based assessments (MRI, General Movements Assessment, Hammersmith Infant Neurological Examination; 1-way analysis of variance, P < .05), with infants who received more assessments tending to be referred at younger ages. There were no significant differences in age at first referral or in age at referral to early intervention among those receiving different numbers of assessments.

Table 3.

Comparison of Referrals to Outpatient Therapies and Early Intervention Between Earlier (2010–2017) and Later (2018–2022) Cohorts.

Diagnosis year	Therapy type	% referred at or before diagnosis	Mean corrected age (mo) ± SD of referral
2010–2017	Outpatient	61	8.7 ± 8.0*
	Early intervention	98	1.5 ± 1.8
2018–2022	Outpatient	92	4.4 ± 4.2*
	Early intervention	77	4.3 ± 5.0

^*t test P value <.05.

Figure 3.

Age at referral to rehabilitation therapies (early intervention and outpatient). (A) Corrected age at referral to outpatient therapy (physical therapy, occupational therapy, or speech/language pathology) by year of diagnosis (r = −0.435, P < .001, n = 66). (B) Corrected age at referral to Early Intervention by year of diagnosis (r = 0.330, P = .009, n = 62). Gray lines represent linear regression fit lines.

Discussion

In this retrospective medical record review, we analyzed diagnosis and intervention referral practices in a cohort of children that attended a high-risk infant follow-up clinic in Madison, Wisconsin, and received a cerebral palsy diagnosis between 2010 and 2022. We found that the average age at cerebral palsy first mention decreased significantly following the publication of early detection guidelines by Novak et al. in 2017, whereas the average age at cerebral palsy diagnosis did not significantly differ over time. Both age at cerebral palsy first mention and age at cerebral palsy diagnosis varied with the number of evidence-based assessments implemented, with lower ages at first mention and diagnosis associated with increased use of evidence-based assessments. All but one of the infants in the sample were referred to rehabilitation therapies at or before diagnosis, and a change in referral patterns was noted with earlier referral to outpatient therapies over time.

Age at cerebral palsy first mention decreased significantly between the earlier and later cohorts, although age at cerebral palsy diagnosis did not. Prior qualitative literature regarding parent experiences suggests that ongoing communication regarding the possibility of a cerebral palsy diagnosis and providing an interim high-risk diagnosis is beneficial to parents, allowing them to seek information and treatment to support their child’s development.² Furthermore, a “high-risk” designation may expedite access to early intervention.¹⁴ Providers may prefer to inform families that their child is at high risk for cerebral palsy and then wait until a subsequent follow-up visit to make a diagnosis, allowing parents time to adjust to the diagnosis and facilitating an active diagnosis conversation. This may be one reason why age at cerebral palsy diagnosis showed little change over time, whereas age at cerebral palsy first mention decreased: providers may have had increased diagnostic certainty at younger ages, but preferred to give families time to understand the diagnosis before officially providing it. Another possible factor associated with the lack of change in cerebral palsy diagnosis could be changes in the patient population at the NBFU clinic (e.g., greater medical complexity). Although formal diagnosis can be beneficial for accessing services, equipment, and caregiver support, an earlier high-risk designation may still allow parents to seek diagnosis-specific support for their child, including outpatient or early intervention therapies. Studies of parent experiences with the diagnosis process have reported that families value having transparent information about the clinician’s differential diagnoses, and that their involvement in the diagnostic process was beneficial.²

The exploratory regression analyses for age at cerebral palsy first mention and age at cerebral palsy diagnosis further illuminate the complexities of these timelines. Corrected age at the first visit to the NBFU Clinic was a significant predictor for cerebral palsy first mention, potentially because infants who presented to the NBFU Clinic after 5 months were typically not assessed with the General Movements Assessment and may have had fewer newborn-detectable risk factors. Number of assessments was not a significant predictor in the exploratory model of age at cerebral palsy first mention, although finding pathology on an MRI scan was, highlighting the key role of neuroimaging in supporting a high-risk designation for cerebral palsy. Additionally, cerebral palsy distribution was a significant predictor of age at cerebral palsy first mention in the exploratory regression model, but was not a significant predictor of age at diagnosis. In the exploratory regression model, post hoc testing identified that diparetic cerebral palsy was significantly associated with younger age at cerebral palsy first mention compared with hemiparetic cerebral palsy; however, quadriparetic cerebral palsy was not significantly different from hemiparetic cerebral palsy after adjusting for other model variables. Previous studies have found that children with spastic quadriparetic cerebral palsy tend to be diagnosed earlier than children with other types of cerebral palsy, although this may be influenced by motor severity.^17–19 The model for cerebral palsy diagnosis included a variety of factors as significant predictors: number of assessments (among MRI, General Movements Assessment, Hammersmith Infant Neurological Examination), year of diagnosis, diagnosing provider, and pathology indicated on MRI. Both models explained less than 30% of the variance in the respective outcomes (cerebral palsy first mention and cerebral palsy diagnosis), indicating the high variability of diagnosis experience and the likely existence of other individual patient, provider, and external factors that influence cerebral palsy high-risk designation and diagnosis time frames.

A study evaluating the implementation of the 2017 early detection guidelines in a high-risk infant follow-up network in the United States found a decrease in average infant age at diagnosis from 19.5 to 9.5 months following implementation efforts.¹⁰ In the present study, the average age at diagnosis remained around 15 months in the earlier and later cohorts. In comparison, the average age at cerebral palsy first mention decreased from approximately 11 to 6 months. Barriers that have been identified to the implementation of evidence-based assessment pathways include lack of provider training on standardized assessments, the time- and resource-intensive nature of obtaining training, and the feasibility of obtaining neuroimaging due to cost or clinical instability.^10,11,20,21 In this sample, Hammersmith Infant Neurological Examination utilization was low compared to other recent studies of guideline implementation for early detection of cerebral palsy.^14,15 In some circumstances, a subset of the Hammersmith Infant Neurological Examination or a nonstandardized neurologic examination was completed. Although the NBFU clinic provided dedicated funds and time for General Movements Assessment training, the Hammersmith Infant Neurological Examination did not have a specific training or implementation process. Increasing provider competency and comfort in performing the Hammersmith Infant Neurological Examination could further improve diagnosis timelines in this setting.²² When the General Movements Assessment or Hammersmith Infant Neurological Examination is unavailable, our data suggest that using other assessments (eg, Alberta Infant Motor Scale, Test of Infant Motor Performance, or Developmental Assessment of Young Children) recommended by the guidelines may be beneficial toward an earlier high-risk designation and diagnosis.

Additionally, it is important to note that not all children seen in the NBFU Clinic had newborn-detectable risk factors, and 32% of children in this study were seen for a first visit in the follow-up clinic at greater than 6 months of age. This number likely includes children who did not have newborn-detectable risk factors and those for whom referral to specialty care was delayed. Outreach educational efforts for referral pathways have been found to increase the referral of infants with newborn-detectable risks to follow-up programs.¹¹ Diagnosis timeline may also be impacted by NBFU Clinic appointment attendance and scheduling. Attendance may be affected by many factors, including the distance from the clinic, degree of parental support and resources, and transportation access.^23,24

Overall, the age of referral to therapies was lower than the age of diagnosis of cerebral palsy. This aligns with a recent international clinical practice guideline that advocates for beginning intervention at the time of suspected diagnosis to capitalize on a critical period of neuroplasticity instead of a “wait and see” approach.⁵ Subsequently, early diagnosis may promote access to cerebral palsy–specific therapies and resources. One study reported that children who were referred by a neonatologist or enrolled in neonatal follow-up were referred to physical therapy services significantly earlier than children who were not receiving this specialty care; these children were also more likely to be born preterm and had a greater number of neonatal risk factors.²⁵ A study in Canada of children born between 2008 and 2011 found a similar trend, whereby children who were referred for diagnosis by a primary care practitioner were referred to therapy services approximately 6 months later than those referred for diagnosis by a medical specialist.¹⁷ In our study, 44% of children with a known referral date to early intervention were referred at <1 month of age, commonly at neonatal intensive care unit discharge. However, intervention referral timing should not be generalized to individuals with cerebral palsy who did not receive care in the neonatal intensive care unit or participate in a high-risk neonatal follow-up program. Furthermore, referral practices were noted to shift, with earlier referrals to outpatient services and later referrals to early intervention over time. Outpatient pediatric therapies became more available throughout the state of Wisconsin during this period, and providers may have increasingly identified a need for intensive services in children at risk for cerebral palsy. Given the wide range in therapy service availability and early intervention delivery models across the United States, this observed shift in referral practice may not be generalizable to other health systems or regions.

The time frame of this study overlapped with the COVID-19 pandemic, which had many consequences for health care. At the NBFU Clinic, visits were conducted fully via telehealth (phone or video) from March to July 2020, which may have impacted appointment timing and assessment administration, and presented challenges for provider confidence in diagnosis. As early intervention and many outpatient therapies also transitioned to virtual care, some families may have had more limited access to services or chosen to delay access during the pandemic. In surveys, caregivers of children with disabilities reported decreased access to rehabilitation therapies during the early months of the pandemic, as well as reduced satisfaction with care provision.²⁶ One study of a newborn follow-up clinic during the COVID-19 pandemic found comparable intervention referrals generated in telemedicine vs in-person visits; the generalizability of these findings is unknown.²⁷ In 2020, the Wisconsin Department of Health Services reported a decrease in referrals and enrollment in early intervention programs, which was believed to be impacted by the pandemic.²⁸ After a concerted Child Find campaign to inform families and providers about early intervention services, enrollment returned to baseline in 2021.²⁸ COVID-19 may have also exacerbated disparities in appointment attendance, diagnosis, and therapy access based on factors like socioeconomic status or medical complexity.^29,30 Further research to assess potential changes in diagnostic and referral practices as health care systems emerge from the pandemic may illuminate lasting changes in care practices.

Our study has several limitations related to the population included and the retrospective nature of medical record access. This study only included patients who had access to care at a high-risk infant follow-up clinic. Patients at the NBFU Clinic primarily come from South/Central Wisconsin and Northern Illinois, and families report a range of urban and rural home locations, insurance payers, and socioeconomic status. However, patients seen at the NBFU Clinic may not represent the general population of children with cerebral palsy in Wisconsin or the United States. Our sample included a somewhat larger population with spastic or mixed quadriparesis than seen in register reports (e.g., the 2023 Australian Cerebral Palsy Register Report).³¹ A potential reason for the higher percentage could be that children with greater medical complexity or functional impairment were more frequently referred to the NBFU for specialty follow-up care. As the analysis was limited to records within the UW-Health system, dates of first cerebral palsy mention and therapy referrals that occurred outside the system could not be included in the analysis. Although a documented diagnosis within the medical record was an inclusion criterion for the study, this does not exclude the possibility that another provider could have diagnosed a child with cerebral palsy and there was a subsequent rediagnosis by a UW-Health provider. Furthermore, children may have received evidence-based assessments from a provider outside of the UW-Health system; however, if those assessments were not available to the diagnosing provider, they are unlikely to have influenced the diagnosis timeline. Additionally, we did not complete longitudinal medical record review after the diagnosis to determine whether any of the diagnoses were false positives (e.g., children who later had their cerebral palsy diagnosis removed).

Finally, this study included dates of therapy referral as a proxy for therapy access because of the greater consistency of documentation in the medical record. Families may not have initiated or continued therapies to which they were referred, indicating that our numbers may overrepresent actual therapy access. Furthermore, therapy access may have been delayed for various reasons (eg, long wait times, family availability, medical status). Therefore, referral dates may not be an accurate metric to determine the timing of therapy initiation. We also do not report information about the frequencies or content of outpatient or early intervention therapies. Future studies are needed to inform the relationship between early diagnosis and dosing/content of rehabilitation therapies and to assess outcomes related to access to cerebral palsy-specific therapies.

Conclusion

Evidence-based, cerebral palsy–specific assessments were increasingly integrated following the publication of early detection guidelines by Novak et al. in 2017. Infants who received multiple evidence-based assessments were diagnosed at a younger age than those who did not, underscoring the importance of evidence-based assessments in achieving earlier diagnoses of cerebral palsy. Predictors identified in regression models explained less than 30% of the variability in cerebral palsy diagnosis timelines, suggesting that many individual patient, provider, and external factors impact diagnosis pathways. Nearly all infants were referred to 1 or more rehabilitation therapy services before diagnosis, which aligns with current guidelines supporting proactive intervention strategies. Age at referral to outpatient therapies decreased over time, whereas age at referral to early intervention increased, potentially reflecting changes in referral practices over time. As clinical settings implement early diagnosis pathways and assessment tools for cerebral palsy, assessing the subsequent impact on high-risk designation and diagnosis timelines for infants at risk for cerebral palsy is important to identify and address remaining gaps in guideline implementation. Further research is needed to clarify the relationships between early diagnosis, access to cerebral palsy-specific rehabilitation therapies, and child outcomes.