concurring in part and dissenting in part.
I agree with the majority that Casag-rande has provided enough evidence to create a genuine issue of material fact regarding the delay in his reinstatement, but in my view this was not enough to conclude that he is entitled to judgment on this issue as a matter of law. With respect *503to the retaliation issue, the district court properly ruled for the defendant.
I.
On the interference claim, record evidence supports a conclusion that Casag-rande was entitled to be reinstated to his job in January of 2013. It is undisputed that Casagrande’s leave qualified for FMLA protection in December of 2012. Further, Casagrande informed OhioHealth of his desire and ability to return to work on January 18, 2013. It is also undisputed that OhioHealth did not reinstate Casagrande to his position on 7 Orange until March of 2013. Employees able to return from FMLA-eligible leave are entitled, under the FMLA, to be reinstated to their prior position or an equivalent position without delay. See 29 U.g.C. § 2614(a)(1)(A)-(B); Hoge v. Honda of America Mfg., Inc., 384 F.3d 238, 249 (6th Cir. 2004). Although the defendants argue that any delay was due to Casagrande’s failure to provide a return-to-work release per OhioHealth’s policy, there is evidence raising a dispute over whether such a policy existed. To be- sure, an employer can have a policy requiring a return-to-work release1 before reinstating an employee on FMLA leave. 29 C.F.R. § 825.312(a). But this policy must be uniformly applied, id. and Casagrande has elicited sufficient evidence to dispute whether OhioHealth applied its policy uniformly. OhioHealth indeed sent several letters to Casagrande that stated a requirement for him to provide a return-to-work release prior to being reinstated. OhioHealth’s actions, however, conflict with this supposed policy. Not only did OhioHealth let Casagrande return from his previous medical leave without providing a release, but at least one OhioHealth official also discouraged Casagrande from providing a release after he stated his ability to return to work. This contradicts OhioHealth’s assertion that the delay in failing to provide a release was Casagrande’s fault. Because there is á dispute over whether Casag-rande had to provide a return-to-work release prior to being reinstated, there is a dispute over whether he was entitled to reinstatement in January 2013. The district court thus erred in concluding otherwise.
Casagrande has made the remaining necessary showings for his interference claim. Other than the entitlement element analyzed above, the defendants on appeal do not dispute that Casagrande has satisfied the elements of his interference claim. What the defendants do dispute,' however, is whether Casagrande was harmed by the delay. The district court held that Casag-rande could not show harm—a requirement for an interference claim, Edgar v. JAC Products, Inc., 443 F.3d 501, 507-08 (6th Cir. 2006)—because OhioHealth gave him a check for back wages for the twenty-day delay between when he submitted his return-to-work release in February and when he was actually reinstated in March. On appeal, the defendants make the same argument. It is not disputed that the offered wages cover only the February— March delay. OhioHealth did not offer Casagrande any back wages for January even though, if that was the start date of OhioHealth’s FMLA interference, that would be when Casagrande’s harm began. Thus, because the offer for back wages covered only the delay from February to March and did not cover January, Casag-rande has elicited sufficient evidence to satisfy the harm inquiry, and the district court erred in awarding summary judg*504ment to the defendants on the interference claim.
Assuming that Casagrande has asserted a prima facie case with regard to retaliation, the defendants had a legitimate, non-retaliatory reason for firing him.2 Ohio-Health’s reasons for firing Casagrande were patient-safety concerns and a failure to follow nursing policy. These are certainly legitimate reasons that on their face have nothing to do with retaliation. Although Casagrande states in a heading in his brief that OhioHealth cannot provide a legitimate, non-retaliatory reason for terminating his employment, he provides no evidence, case citations, or further analysis of this issue. Thus, the defendants have shown that there was a non-retaliatory basis for Casagrande’s termination.
The record evidence supports none of Casagrande’s theories as to pretext. There are three ways a plaintiff can show pretext: “showing that the employer’s proffered reasons (1) have no basis in fact; (2) did not actually motivate the action; or (3) were insufficient to warrant the action.” Seeger v. Cincinnati Bell Telephone Co,, 681 F.3d 274, 285 (6th Cir. 2012). None of these theories is availing for Casagrande. First, the patient-safety and following-policy concerns that motivated the defendants’ termination of Casagrande had bases in fact. Casagrande admitted in his deposition that most of the alleged performance issues occurred, though he attempted to explain the reasons for them or mitigating circumstances. The sheer number of performance concerns gave OhioHealth a factual basis for ending Casagrande’s employment. Furthermore, OhioHealth undertook reasonable investigation prior to ending Casagrande’s employment, as evidenced by the fact that it decided not to fire him in April after hearing his version of events and that, in October, it investigated the allegations against him and determined that two of the four performance issues for which he had been placed on fact-gathering suspension were unmerited. Casag-rande has not shown that OhioHealth’s proffered rationale for firing him had no basis in fact. For these same reasons, Cas-agrande has not shown, under the third theory of pretext, that OhioHealth’s rationale was insufficient to justify his termination.
Casagrande has also failed to elicit sufficient evidence to create a dispute of material fact over whether the performance concerns actually motivated his termination. Casagrande has pointed to no direct evidence in the record. He does, however, point to the deposition testimony of two of his coworkers, Sherry Pappas and Tamara Fluty, to establish pretext. But, as the district court held, all of Pappas’s and Fluty’s testimony is either based on hearsay or without personal knowledge of the heightened scrutiny. Pappas testified that “it did seem to [her] that [Sayers] was treating [Casagrande] unfairly,” but this was in the context of her conversations with Casagrande. Thus, although what Casagrande relayed to Pappas might have sounded unfair, that is not sufficient evidence to support this pretext theory; indeed, it is no better than Casagrande’s bare allegations. Moreover, although Pap-pas stated that Casagrande “was being written up for things that anybody could have done,” this was a direct quote taken from Casagrande in their conversations. Pappas never stated that she personally observed any heightened scrutiny as to Casagrande’s performance. And Fluty pro*505vided much of the same. Fluty testified that after Casagrande returned from his FMLA-eligible leave, Sayers called her one night to investigate a report about Casagrande. In this call, Sayers simply reminded Fluty of her “job as a charge nurse to report” any problems that occur and that Sayers expected Fluty to do her job. In response to a question as to whether she felt pressured by Sayers, Fluty stated “I don’t know. I guess in the beginning. But, you know, I don’t know.” This contradictory testimony does not create a dispute as to pretext. Furthermore, although Fluty stated that Sayers might have been “nitpicking” Casagrande’s performance, she admitted that this was hearsay, even saying “I didn’t experience [i.e., observe] any of [the alleged scrutiny].” “Hearsay evidence may not be considered on summary judgment.” Jacklyn v. Schering-Plough Healthcare Products Sales Corp., 176 F.3d 921, 927 (6th Cir. 1999). Because Casagrande has elicited no admissible evidence that demonstrates heightened scrutiny as to his job performance after his return from FMLA leave, he cannot avail himself of the second theory of pretext.
Without any admissible evidence showing that the defendants’ legitimate reason for his termination was not the actual one, Casagrande’s retaliation claim fails. He attempts to argue that the “changing justifications” for his termination are sufficient to show pretext. That the defendants ultimately based their termination of Casag-rande on two incidents instead of the four incidents originally stated does not amount to a “changing justification.” As Casagrande admits, the defendants investigated the incidents and simply removed, the two that were least supported. Removing two incidents from a four-incident list does not show that the remaining reasons or underlying concerns were pretextual. Moreover, Casagrande admitted in his deposition that most of the alleged performance issues occurred, though he attempted to explain the reasons for them or mitigating circumstances. The “changing justifications” argument therefore is without merit. Because Casagrande has made no other arguments on appeal regarding pretext, and because an argument not made on appeal is waived, City of Cleveland v. Ameriquest Mortgage Securities, Inc., 615 F.3d 496, 501 n.5 (6th Cir. 2010), the district court’s award of summary judgment to the defendants on the retaliation claim should stand.3
III.
For the foregoing reasons, I would affirm the district court’s grant of summary judgment as to the retaliation claim and reverse as to the interference claim.
. The FMLA regulations use the term "fitness-for-duty certification,” 29 C.F.R. § 825.312, and OhioHealth uses “return-to-work release," but these terms are interchangeable.
. FMLA retaliation claims fall under the familiar McDonnell Douglas framework. Edgar v. JAC Products, Inc., 443 F.3d 501, 508 (6th Cir. 2006).
. Furthermore, an inference of pretext does not follow from evidence that Casagrande had only one documented performance issue before his first leave of absence, but then received two written reprimands after returning from his first leave of absence, as well as several verbal reprimands and a written warning after returning from his second leave. Without Pappas’s and Fluty’s inadmissible testimony, the evidence in the record shows only that Casagrande was disciplined more often in the months leading up to his termination. Even viewed in the light most favorable to Casagrande, this evidence simply does not show that OhioHealth terminated Casagrande for any reason other than Casag-rande’s disregard for patient safety and failure to follow nursing policy.